Вакансии

Вы будете создавать и внедрять решения на базе генеративного ИИ: мультиагентные системы, RAG-конвейеры, интеграции с внутренними сервисами и автоматизацию бизнес-процессов. Строить и сопровождать RAG-решения: векторные хранилища, пайплайны, ретривал.

Задачи: — трансляция бизнес-потребностей в технические требования к функционалу и логике AI-агентов — проектирование интеграций: описание взаимодействия агентов с внешними API (входы/выходы, форматы данных, обработка ошибок, таймауты, retry-логика) — проектирование меж-агентского взаимодействия: оркестрация агентов, передача контекста, приоритизация запросов и сценарии эскалации — сопровождение до ПРОМ: совместно с промпт-разработчиками доводите требования до готовых промтов, контролируете соответствие промтов исходным бизнес-требованиям и передача системных требований в команду сопровождения — взаимодействие с заказчиками, менеджментом и командой разработки, обеспечение прозрачности процесса — исследование рынка и проверка гипотез: быстрая оценка применимости AI-решений под задачу, сравнение подходов (LLM, RAG, инструменты, мультиагентность), подготовка обоснованных рекомендаций Требования: — уверенный технический бэкграунд: понимание архитектуры веб-систем, REST API, gRPC, форматов данных (JSON, OpenAPI/Swagger, Protobuf), интеграционных паттернов (sync/async, webhooks, очереди) — понимание AI-концепций на уровне теории либо применения: как работают LLM, промпт-инжиниринг, ограничения и «галлюцинации»; RAG (retrieval-augmented generation), векторные БД; архитектура AI-агентов: инструменты (tools/function calling), оркестрация, мультиагентные сценарии; знакомство с LangGraph/LangChain или аналогичными фреймворками — понимание, что такое harness (core-платформа для GenAI-агентов) и его составляющих — умение проектировать логику агентов: сценарии, состояния, ветвления, обработка ошибок, fallback-сценарии — навыки декомпозиции задач и написания точных, проверяемых технических требований (User Story + критерии приёмки, или спецификации) — опыт взаимодействия с бизнесом и перевода требований в технические спецификации — опыт работы с PostgreSQL/SQL (чтение схем, описание маппинга данных) — умение быстро проверять гипотезы: исследовать рынок AI-решений, оценивать применимость технологии под задачу Будет плюсом: — понимание Java/React-стека на уровне общения с командой (что реально, а что нет в существующей архитектуре) — опыт в продуктовых командах, где аналитик сопровождает фичу от идеи до релиза — опыт написания технических заданий для промпт-разработчиков — базовые навыки прототипирования (Python/псевдокод/конфигурация) для быстрой проверки гипотез Условия: — комфортный современный офис — корпоративный спортзал и зоны отдыха — уникальная система обучения Сбера для профессионального и карьерного развития — программа адаптации и помощь руководителя на старте — расширенный ДМС и льготное страхование для семьи — гибкий дисконт по ипотечному кредиту, равный 1/3 ключевой ставки ЦБ — подписка Прайм с возможностью совместного использования на трёх близких — вознаграждение за рекомендацию друзей в команду Сбера — корпоративная пенсионная программа.

Задачи: — Проектировать и создавать AI-агентов и решения на базе больших языковых моделей (LLM) — Разрабатывать конвейеры данных и lakehouse-решения в Databricks — Интегрировать AI-агентов с Databricks и другими платформами данных — Создавать решения на Python для сбора данных, веб-скрапинга и индексирования контента — Вносить вклад в архитектуру и технические решения — Тесно сотрудничать с международной инженерной командой Требования: — Опыт работы на уровне senior в области разработки ПО или AI-инженерии — Уверенное владение Python — Практический опыт работы с GenAI, LLM и Agentic AI — Опыт работы с инженерией данных и конвейерами данных — Практический опыт работы с Databricks — Способность брать ответственность и участвовать в архитектурных решениях — Владение английским языком на уровне C1 Будет плюсом: — Знание графов знаний и онтологий — Опыт веб-скрапинга — Знание Databricks Genie — Опыт работы с агентскими фреймворками и разработкой с нуля архитектуры Условия: — Медицинская страховка — Более 21 дня отпуска — Возможности профессионального развития — Поддержка карьерного коуча Проект международный, европейское сотрудничество, локация Ташкент, Узбекистан.

Чему вы научитесь: • Исследование угроз GenAI: создание, поддержка и развитие инструментов для анализа и поиска угроз; • DevOps: CI/CD, контроль версий (git), тестирование качества, инфраструктура для работы (linux); • Внедрение практик AI PDLC: исследование и применение методов и инструментов создания и развития программного обеспечения с применением передовых ИИ-моделей; • Погружение в экосистему: изучение и практическое применение передовых фреймворков и библиотек для работы с LLM и AI-агентами; • Управление командой: формирование и развитие исследовательской команды (3–5 человек), приоритизация исследований, защита решений перед руководством и смежными командами; • CTF и хакатоны: формирование и капитанство команды на международных CTF по AI/ML security, трансляция найденных техник в работу направления; • Публикации и нормативные документы: оформление результатов исследований, вклад в нормативные документы безопасности AI. Требования: • Профиль и образование: законченное высшее образование в области информационных технологий, компьютерных наук, прикладной математики, программной инженерии или смежных направлений • Продвинутые навыки программирования: понимание принципов проектирования и разработки программного обеспечения, опыт в написании и чтении программного кода (понятный, поддерживаемый) • Навыки DevOps: работа с git, администрирование linux, автоматизация этапов жизненного цикла разработки ПО • LLM и AI-агенты: экспертный уровень применения LLM для разработки ПО, глубокое понимание архитектуры агентов и их векторов атак • Английский язык на уровне не менее чтения технической документации, статей и участия в международных CTF • Опыт командной работы в проектах разработки: канбан (спринты, трекинг задач), jira/git/confluence, презентации и умение понятно объяснять свои идеи команде/руководству и согласовывать требования • Adversarial ML и red teaming LLM: практический опыт поиска уязвимостей в GenAI-моделях (prompt injection, jailbreaking, tool poisoning, data extraction) • Знание инструментария для LLM security: Giskard, Garak, LLM Guard и др. • CTF и хакатоны: участие в международных CTF по AI/ML security (желательно — победы или призовые места).

Location: Москва Наша команда отвечает за **независимую оценку и контроль качества генеративных моделей Банка.** Наша основная задача оценить флагманские генеративные модели Сбера – GigaChat, Kandinsky, GigaCode и PhysicalAI решений. Мы первыми видим, на что способна новая модель и проводим оценку новых возможностей прежде чем она выйдет к пользователям. Мы не просто «проверяем метрики», а: * глубоко разбираемся в подходах команд-разработчиков и в бизнес-процессах, где работают модели * критически рассматриваем архитектурные и методологические решения и предлагаем улучшения * разрабатываем собственные проверки, метрики и бенчмарки и фреймворки для их запуска * проводим оценку логов пользователей для построения наиболее релевантного тестирования Отдельный фокус – R&D деятельность: * создание новых методологий оценки качества * участие в разработке и развитии передовых банковских и российских бенчмарков (MERA, POLLUX и др.) * адаптация передовых мировых бенчмарков под специфику банковских кейсов * и многое другое Эта роль про независимое измерение фундаментального качества модели: не про их обучение – этим занимается отдельная команда Обязанности **Придумывать, как вообще измерить модель** ИИ стремительно развивается в мире и выходят новые модели и их виды, однако не всегда любая новая модель стала лучше предыдущей. Вам будет необходимо создать план по её проверке: что именно измерять, на какой выборке, с чем сравнивать, какая разница вообще будет считаться разницей в реальном качестве. **Отвечать за достоверность измерений** Полученные метрики не всегда отражают только лишь качество модели. Нужно уметь отделять реальный прирост качества от прочих факторов: ловить утечки данных в обучение, отличать дефект модели от дефекта замера, промпта или обвязки и проявлять тот же скепсис к собственной методологии: измеряем ли мы то, что заявили. **Разбираться в причинах ошибок** «Модель ошиблась» – это не вывод, а начало работы. Найденная вами причина той или иной проблемы модели превращается в рекомендацию команде разработки, а рекомендация – в следующую версию модели, которая этой ошибки уже не делает. **Строить автоматическую оценку** Разрабатывать LLM- и VLM-судей под конкретные задачи, разрабатывать с нуля собственные или дорабатывать существующие библиотеки валидации полного цикла и следим за фронтиром AI-исследований в области оценки моделей. **Строить то, чем пользуются другие** Фреймворки оценки, который вы развиваете – общие для команд по всему Банку. Бенчмарки, которые вы делаете, являются достоверной базой реального качества для разработчиков Банка и вне. Требования * **Умение проводить технический и научный ресерч**: находить и критически разбирать актуальные статьи, бенчмарки, модели и реализации; переводить ресерч в конкретные идеи для тестирования и развития внутренних подходов. * **Статистика и дизайн эксперимента:** доверительные интервалы, размер выборки, проверка гипотез, поправки на множественные сравнения * **Понимание современных AI-агентов и агентных систем**: понимать, чем агент отличается от обычного вызова LLM; знать основные компоненты агентной системы: роли, инструменты, контекст, память и оркестрация; * **Понимаете, как устроены генеративным модели изнутри:** трансформер и механизм внимания, диффузии, токенизация, параметры генерации, этапы обучения * **Опыт с генеративными моделями и понимание, как их оценивают:** метрики, бенчмарки, human evaluation, Model-as-a-judge и его ограничения * **Продуктовое мышление**: связывать тестирование с реальными пользовательскими сценариями и понимать, какие ошибки критичны для продукта, понимать влияние результатов на релиз и дальнейшее развитие модели. * **Python на продакшн-уровне:** чистый код по PEP 8, ООП и типовые паттерны проектирования, аннотации типов, привычка к линтерам и статическому анализу (ruff, mypy). Создаете и поддерживаете production-код eval-пайплайнов, читаете и ревьюите чужой. **Будет плюсом:** R&D-опыт или участие в создании бенчмарков; мультимодальные модели (изображения, видео, аудио); evaluation-фреймворки и harness; RAG; работа с асессорами; coding-агенты (Claude Code, Codex, Cursor); опыт работы с оркестрацией: LangGraph, OpenClaw, навыки вайбкодинга и AI-assisted coding **Стек:** Python, pandas, NumPy, PyTorch, HuggingFace (transformers, datasets), vLLM, SGLang, инференс-API, S3, Git, Docker, CI/CD, Linux, Bash. Условия * комфортный современный офис рядом с м. Кутузовская * гибкий формат и удаленной работы, в т.ч. на территории РФ * ежегодный пересмотр зарплаты и годовая премия * корпоративный спортзал и зоны отдыха * более 400 образовательных программ СберУниверситета для профессионального и карьерного развития * расширенный ДМС, льготное страхование для семьи и корпоративная пенсионная программа * гибкий дисконт по ипотечному кредиту, равный 1/3 ключевой ставки ЦБ * подписка Прайм с возможностью совместного использования на трёх близких * скидки на продукты компаний-партнеров * вознаграждение за рекомендацию друзей в команду Сбера.

Задачи: — Проектирование и разработка микросервисной архитектуры на Java 17+ с использованием Spring Boot (MVC, Data JPA, Security) — Построение пайплайнов Retrieval-Augmented Generation (RAG): от парсинга сырых данных до индексации эмбеддингов — Разработка и поддержка векторных баз данных Qdrant: проектирование схем шардирования, настройка гибридного поиска (BM25 + Dense Vectors) и оптимизация запросов — Создание мультиагентных систем на базе LangChain/LangGraph для автоматизации бизнес-процессов — Глубокий промт-инжиниринг: написание промптов под разные сценарии одного агента, техники Few-shot, Chain-of-Thought, структурированный вывод (JSON mode) — Отладка семантики моделей: снижение энтропии ответов, управление вариативностью через параметры инференса, механизмы self-healing при деградации внешних провайдеров LLM — Оптимизация производительности PostgreSQL: тюнинг сложных SQL-запросов, работа с индексами (B-tree, GIN/GIST, триграммы), проектирование транзакционной изоляции — Вынос вычислений в контейнерную среду: Docker, Kubernetes (Helm), построение отказоустойчивых CI/CD-конвейеров в GitLab CI/Jenkins Требования: — Уверенное владение Java 17+, понимание многопоточности (CompletableFuture, виртуальные потоки Project Loom — будет плюсом) — Экспертный уровень владения стеком Spring: жизненный цикл бинов, аспектно-ориентированное программирование, тонкая настройка пулов соединений БД и параметров безопасности — Опыт проектирования реляционных моделей в PostgreSQL, умение читать EXPLAIN ANALYZE, эффективные миграции (Liquibase/Flyway) — Практический опыт работы с векторными СУБД (Qdrant) или Weaviate/Pinecone/Milvus, понимание метрик близости (Cosine Similarity, Dot Product, Euclidean Distance) — Опыт разработки с библиотеками LangChain и/или LangGraph, построение графов состояний, интеграция внешних инструментов (tool calling) — Навык написания эффективных промптов (Prompt Engineering) и управления поведением LLM без изменения весов модели — Опыт работы с Docker и Kubernetes (манифесты, Helm-чарты), знание Linux shell на уровне уверенного пользователя — Настроенные процессы CI/CD (GitLab CI или Jenkins): сборка Maven/Gradle, покрытие тестами (JUnit 5, Testcontainers), статический анализ кода

Location: Colombia; Costa Rica Workplace: remote Provectus is a Premier AWS Partner and an Anthropic partner, featured in Forrester's AI Technical Services Landscape. We are roughly 420 engineers building production AI for enterprise clients in Financial Services, Healthcare and Life Sciences, Retail, and Entertainment. We build custom systems on live client data, not off the shelf products and not demos. We are opening a Solution Owner seat on our Delivery team in Colombia and Costa Rica. This is an early career seat, but it is not a training seat. You will sit between the client and the engineering team on a real engagement, own a slice of scope, and be expected to figure out the next step yourself. The profile we are looking for is specific: you started technical, as a software, data, or ML engineer, and you have already moved toward the business side of building products, or you are ready to make that move now. You are the person on the team who automates their own work with AI tools before anyone asks. What You Will Do: Own a defined slice of a client engagement, from use case framing through backlog, release, and result Run discovery sessions and workshops with client stakeholders, and turn what you hear into written scope engineers can build against Translate technical constraints into business language, and business goals into technical requirements Maintain a groomed backlog with sized, technically specific stories, and drive sprint planning and release cadence Prototype rapidly on live client data using our delivery stack, including Claude, Claude Code, MCP servers, and Skills Work alongside GenAI and agentic AI builds, including RAG, agent workflows, and LLM integrations, closely enough to judge quality and explain trade offs Define success metrics for your slice of scope and report progress honestly, including when something is not working Build reusable artifacts, demos, and documentation so the next engagement starts faster than yours did Partner with Principal Architects, Forward Deployed Engineers, and account teams across US and LATAM time zones What You Will Bring: 2 to 4 years of total professional experience, including hands on engineering time as a software, data, or ML engineer At least 1 year in a product, delivery, or client facing role such as Product Owner, Technical Product Owner, Business Analyst, Delivery Lead, or equivalent Daily use of GenAI tools in your own workflow, Claude and Claude Code preferred, with concrete examples of workflows you redesigned Ability to read code and technical documentation well enough to hold a real conversation with senior engineers Practical Agile and Scrum experience: backlog, sprints, planning, retrospectives Written and spoken English at B2 or above, comfortable leading a client call without a manager in the room Self direction as a default. You are given a goal and ambiguity, not a task list, and you move anyway Based in Colombia or Costa Rica Nice to Have: Exposure to Financial Services and Insurance or Healthcare and Life Sciences clients Working understanding of GenAI concepts: prompt engineering, RAG, embeddings, vector databases, agentic patterns Familiarity with AWS, especially Bedrock, SageMaker, Lambda, or S3 Experience shipping an LLM or RAG feature that reached real users Experience building your own automations, scripts, MCP integrations, or internal tools Consulting or professional services background Certifications such as CSPO, SAFe PO/PM, or AWS Cloud Practitioner What We Offer: Real client delivery on frontier AI from month one, not shadowing Direct access to Principal Architects and Forward Deployed Engineers as your working peers Anthropic and AWS partnership work, with early access to tooling and premium AI subscriptions Internal training programs and full support for AWS and Claude certifications A clear growth path toward Senior Solution Owner or Solutions Architecture Remote work with flexible hours across US and LATAM overlap Long term B2B collaboration Equipment budget and all the tooling you need

Location: San Francisco Bay Area, California; New York City Metropolitan Area Workplace: remote About the role: Provectus is a Premier AWS partner and an Anthropic Strategic Partner at the forefront of applied AI, helping enterprises turn Claude, agentic systems, and their own data into measurable business outcomes through bespoke applications, managed services, and advisory engagements. With offices in North America, LATAM, and EMEA, we partner with clients worldwide and are obsessed with reimagining how they operate and compete. Our work centers on two verticals — Financial Services & Insurance and Healthcare & Life Sciences — where we deploy five pre-built AI Blueprints: Submission Flow, Portfolio Lens, Asset Flow, Revenue Flow, and Evidence Lens. Each Blueprint rebuilds a critical business process front to back, shipped from working code and tuned to a client's specific book, regulators, and operating posture. We embed engineers and leaders inside client operations as Forward Deployed Engineers (FDE) and Forward Deployed Executives (FDX) — people who learn the work, ship the system, and own the outcome. Our team holds 100+ AWS certifications, is Claude Code certified, and co-delivers Anthropic's Agentic SDLC program, Cowork Activation, and AI Bluprints. You will do the customer’s job before you automate it. Most AI engagements fail the same way: someone gathers requirements, someone writes a PRD, and a team ships a workflow nobody uses. We think the requirements-gathering step is the bug. So we remove it. A Forward Deployed AI Engineer at Provectus spends the first weeks of an engagement in the operator’s seat — as the underwriter, the analyst, the RCM specialist, the claims clinician, whoever actually does the work we’ve been asked to change. You do the job. You learn the constraints from the inside, the ones nobody writes down. Then you sit at a table with that operator and a Forward Deployed Executive and rebuild the function from first principles — and you are the one who builds it. Three things define how you work: Embedded, not engaged . You are part of the customer’s team and inside their process — not a vendor running a project alongside it. Real tasks, not scope . You are not fenced into a siloed deliverable. You go where the operating problem is. Autonomous . Embedded is not staff-augmented. You own the method; nobody hands you a ticket. You won’t start from zero. Provectus builds industry blueprints — working systems that have already shipped for a customer in your industry. Your engagement starts from that baseline, and what you learn in the field goes back into it. That loop is the difference between an outcome and an invoice. You’ll be measured on whether the Business Unit’s number moved — not on hours, not on scope delivered. This is a role for engineers who have led before — as a founder, a CTO, a staff engineer — and who want to stay in the code while owning the outcome. On most days you’ll be the most senior technical person in the room, and you’ll still be the one shipping. Requirements: 8+ years building software , a substantial share of it writing production code you were accountable for. You are hands-on today and intend to stay that way. You will take the operator’s seat. You are genuinely willing to spend weeks doing someone else’s job — claims processing, underwriting, revenue-cycle work — before you write a line of code. Engineers who need to stay in the IDE should not apply. You learn domains fast. Demonstrated ability to become conversant in an unfamiliar business function quickly enough to argue with the people who do it for a living. Shipped GenAI/LLM systems to production — not demos, not notebooks. You’ve handled the parts that get hard after the prototype works. You evaluate. You have built or owned an eval suite for a non-deterministic system, and you can explain what you measured and why. Strong engineering fundamentals — dropped into an unfamiliar codebase or language, you’re productive. Python and/or TypeScript proficiency; depth matters more than stack. Cloud-native delivery on AWS (GCP/Azure a plus): containers, Kubernetes/ECS, IaC, CI/CD, and the operational reality of a system someone else inherits. Credible with senior stakeholders — you can hold a redesign conversation with a BU head and a scoping conversation with a CTO without losing either room. Comfort with ambiguity and ownership. Engagements start underspecified by design. Closing that gap is the job. Solid AI/ML foundations — you understand what the models do well enough to reason about failure modes, not just call the API. Strong hands-on prodcution experience with Claude Code/Cowork. Fluent English, written and spoken. Nice to have: Prior experience as a founder, CTO, or engineering leader who has chosen to return to individual contribution. Real depth in one of our blueprint industries: financial services, insurance, healthcare, asset management . Consulting, professional services, or other embedded customer-facing delivery. Data platform depth: data lakes, warehouses, streaming and real-time analytics, data mesh and data contracts, governance and data quality. MLOps and classical ML: PyTorch, SageMaker, MLflow. Fine-tuning, distillation, or inference/serving optimization. Graph databases (Neo4j, AWS Neptune). IaC depth: AWS CDK, CloudFormation, Terraform. Open-source contributions or public writing on applied AI. What you’ll do: Take the seat Do the operator’s job for two to four weeks at the start of an engagement. Learn the function from inside, not from a requirements doc. Reach working fluency in a new domain — insurance underwriting, healthcare revenue cycle, asset flow — in weeks, not quarters. Sit with the operator and the Forward Deployed Executive and redesign the function from first principles. Discovery, user research, and PRD-writing collapse into one team that re-imagines its own job. You are all three roles. Build Ship production GenAI systems into the customer’s environment — LLM applications, agentic workflows, retrieval and structured-extraction pipelines, and the services around them. Running software, not recommendations. Build the evaluation harness before you build the feature. When the engagement is bound to a business KPI, “it looked good in testing” is not an answer. Define what working means, instrument it, and let the evals drive the design. Write production code across the stack — backend services, data pipelines, and the AI layer. Python and TypeScript are our centre of gravity; we choose tools to fit the customer, not the résumé. Take systems to production on AWS (GCP/Azure where the customer requires it): containerized, observable, and maintainable after we leave. Start from the blueprint, and feed the blueprint. What you learn in the field becomes the baseline the next engagement starts from. Own the outcome Work in a pair with a Forward Deployed Executive who carries the Business Unit’s KPIs. Your work is measured against the same number. Drive adoption. A system the BU routes around has not shipped. Change management is part of the engineering job here, not a phase after it. Be credible with the customer’s engineers, their operators, and their executives — and be willing to disagree with all three. Shape what we commit to before we commit to it. You’ll have the standing to do it, because you’re the one who will build it. What We Offer: Frontier delivery work across Cowork Activation, Agentic SDLC, and Blueprint Activations in Financial Services and Healthcare The chance to shape how leading enterprises adopt AI, from strategy through first deployment A forward-deployed model working in small, senior teams alongside Principal Architects and Forward Deployed Engineers A growing AI delivery practice where you help build the tooling and frameworks, not just use them Remote-friendly culture High-impact role with direct visibility to leadership. Strong earning potential with performance-based bonuses. Opportunity to work with cutting-edge AI and cloud solutions. B2B contract model or full-time model. Unlimited Vacation policy. Generous health, vision, and dental insurance. 401(K) matching plan. OTE range: up to $250k. The salary range is determined through interviews and a review of the education, experience, knowledge, skills, abilities of the applicant, and alignment with market data. How we hire: Intro conversation — the role, your background, what you want to be doing. Two live engineering sessions. Real problems, your own editor. You may use an LLM assistant (ChatGPT, Claude) — how you work now includes these tools. Autocomplete/agentic coding tools are off for these sessions. The redesign session. We hand you an unfamiliar business function and the constraints of the person who performs it. You have to understand the job well enough to rebuild it — then say what you’d build and how you’d know it worked. No LLMs for this one. Team and practice conversation.

Location: North Macedonia; Bosnia and Herzegovina; Czechia; Croatia; Moldova; Bulgaria; Montenegro Workplace: remote Provectus is an AWS Premier Partner and an Anthropic Strategic Partner, working at the frontier of applied AI. We help enterprises turn Claude, agentic systems, and their own data into measurable business outcomes — through bespoke applications, managed services, and advisory engagements. With offices in North America, LATAM, and EMEA, we partner with clients worldwide. Our work centers on two verticals — Financial Services & Insurance and Healthcare & Life Sciences — where we deploy five pre-built AI Blueprints: Submission Flow, Portfolio Lens, Asset Flow, Revenue Flow, and Evidence Lens. Each Blueprint rebuilds a critical business process front to back, shipped from working code and tuned to a client's specific book, regulators, and operating posture. Our team holds 100+ AWS certifications, is Claude Code certified, and co-delivers Anthropic's Agentic SDLC program, Cowork Activation, and AI Blueprint engagements. Where this role sits You will work in a small, senior pod alongside an FDE and an FDX,  Forward Deployed Executives (FDX) own the commercial relationship and the business outcome. Works alongside the client's leadership or C-suite level to move the client's KPIs. Forward Deployed Engineers (FDE) embed with a client, map the client workflow, identify the business problem underneath it, design and build a working AI solution, present to the client, and transfer the knowledge to the client's team. Owns technical direction of the whole solution. Senior AI Engineer. When an FDE comes back from the client with the business problem, you will help turn that into an agentic system that runs in production and will be responsible for evaluation, observability, and guardrails.  You'll have real ownership of components and of the technical decisions inside them. Requirements: Mindset Proactive and self-directed; you push for clarity rather than waiting for a ticket Excellent communication and problem-solving skills Comfort with ambiguity and ownership. B2+ English, comfortable collaborating across distributed, multicultural teams. Technical depth 5+ years in software or ML engineering, with production systems you were accountable for. Solid AI/ML foundations. You understand what the models do well enough to reason about failure modes. Shipped to production LLM applications and agentic workflows — not demos, not POCs, not notebooks. Agentic orchestration: multi-step workflows, graph-based orchestration, tool use, state management, and recovery from partial failure Experience with LLM APIs (Anthropic, AWS Bedrock, or OpenAI) and agent frameworks. Experience building and optimizing RAG systems in production. Strong engineering fundamentals. Full-stack mindset, comfortable across AI, backend development, and cloud infrastructure. Python and/or TypeScript proficiency; depth matters more than stack. Dropped into an unfamiliar codebase, you're productive. Hands-on AWS in production: Bedrock, Bedrock AgentCore, Lambda, ECS, S3, SQS, ECR, or similar. GCP or Azure is a plus. Cloud-native delivery: containers, ECS or Kubernetes, IaC, and CI/CD applied to AI pipelines. You evaluate. You have built or owned an eval suite for a non-deterministic system, and you can explain what you measured, how you produced ground truth, and what gated a release. Model and agent monitoring, drift detection. Cost and latency discipline: model tiering, caching, and the ability to say what a workload costs to run before it runs. Hands-on production experience with the Claude ecosystem —  Claude Code, CLAUDE.md, hooks, skills files.  Spec-driven development — writing the intent, constraints, and acceptance criteria before you let an agent build — is a strong plus. MCP: you can say why an agent would prefer it to a REST integration. Having authored a server is a plus. Nice to Have: Experience in one of the industries: financial services, insurance, healthcare. Consulting, professional services, or other embedded customer-facing delivery. AWS and Claude Code Certifications A2A: you can explain agent-to-agent interoperability CI/CD pipeline experience (GitHub Actions, GitLab CI) Practical experience with one or more use cases from the following: NLP, LLMs, and Recommendation engines. Experience in an additional language (Go, TypeScript, or Rust). Experience with Apache Spark, Apache Airflow, Kafkа Responsibilities: Work in a pair with an FDE and an FDX. Build and ship production GenAI systems into the customer’s environment (cloud-native data, LLM-based, and agentic AI solutions). Build and optimize RAG systems for production use cases Build the evaluation harness before you build the feature. Write production code across the stack — AI, backend services, data pipelines. We choose tools to fit the customer. Integrate AI components into backend services and RESTful APIs Take systems to production on AWS (GCP or Azure where the customer requires it): containerised, CI/CD. Implement LLMOps and AgentOps practices: agent tracing, prompt and version management, cost and latency monitoring, regression testing, drift detection Start from the blueprint, contribute to enablement and handover: clear documentation, runbooks, and pairing with the client engineers who will inherit the system. Feed reusable components and lessons back into the Provectus Blueprints Participate in technical discussions and architectural decisions Conduct model evaluation, improve failure modes you find, optimize model performance, efficiency, and reliability Mentor junior and mid-level AI engineers, conduct code reviews and share knowledge across the team through documentation, presentations, and workshops. What We Offer: The chance to shape how leading enterprises across LATAM, Europe, and North America adopt AI, from strategy through first deployment A forward-deployed model working in small, senior teams alongside FDE and FDX A growing AI delivery practice where you help build the tooling and frameworks, not just use them Remote-friendly culture Internal training programs with full support for Claude, AWS, and other professional certifications, conference attendance Career growth; we actively develop our engineers Access to the latest AI tools and premium subscriptions Long-term B2B collaboration Private medical insurance or a budget for your medical needs Paid sick leave, vacation, and public holidays Equipment and all the tech you need for comfortable, productive work How we hire: Intro conversation. The role, your background and aspirations, tech questions. Technical interview with live engineering sessions. Real problems, your own editor, you may use an LLM assistant HR Interview. Soft skills and expectations HM interview. Tech questions; a live engineering session is also possible

Location: Ukraine; Spain; Slovakia; Serbia; Yerevan; Slovenia; Poland Workplace: remote Provectus is an AWS Premier Partner and an Anthropic Strategic Partner, working at the frontier of applied AI. We help enterprises turn Claude, agentic systems, and their own data into measurable business outcomes — through bespoke applications, managed services, and advisory engagements. With offices in North America, LATAM, and EMEA, we partner with clients worldwide. Our work centers on two verticals — Financial Services & Insurance and Healthcare & Life Sciences — where we deploy five pre-built AI Blueprints: Submission Flow, Portfolio Lens, Asset Flow, Revenue Flow, and Evidence Lens. Each Blueprint rebuilds a critical business process front to back, shipped from working code and tuned to a client's specific book, regulators, and operating posture. Our team holds 100+ AWS certifications, is Claude Code certified, and co-delivers Anthropic's Agentic SDLC program, Cowork Activation, and AI Blueprint engagements. Where this role sits You will work in a small, senior pod alongside an FDE and an FDX,  Forward Deployed Executives (FDX) own the commercial relationship and the business outcome. Works alongside the client's leadership or C-suite level to move the client's KPIs. Forward Deployed Engineers (FDE) embed with a client, map the client workflow, identify the business problem underneath it, design and build a working AI solution, present to the client, and transfer the knowledge to the client's team. Owns technical direction of the whole solution. Senior AI Engineer. When an FDE comes back from the client with the business problem, you will help turn that into an agentic system that runs in production and will be responsible for evaluation, observability, and guardrails.  You'll have real ownership of components and of the technical decisions inside them. Requirements: Mindset Proactive and self-directed; you push for clarity rather than waiting for a ticket Excellent communication and problem-solving skills Comfort with ambiguity and ownership. B2+ English, comfortable collaborating across distributed, multicultural teams. Technical depth 5+ years in software or ML engineering, with production systems you were accountable for. Solid AI/ML foundations. You understand what the models do well enough to reason about failure modes. Shipped to production LLM applications and agentic workflows — not demos, not POCs, not notebooks. Agentic orchestration: multi-step workflows, graph-based orchestration, tool use, state management, and recovery from partial failure Experience with LLM APIs (Anthropic, AWS Bedrock, or OpenAI) and agent frameworks. Experience building and optimizing RAG systems in production. Strong engineering fundamentals. Full-stack mindset, comfortable across AI, backend development, and cloud infrastructure. Python and/or TypeScript proficiency; depth matters more than stack. Dropped into an unfamiliar codebase, you're productive. Hands-on AWS in production: Bedrock, Bedrock AgentCore, Lambda, ECS, S3, SQS, ECR, or similar. GCP or Azure is a plus. Cloud-native delivery: containers, ECS or Kubernetes, IaC, and CI/CD applied to AI pipelines. You evaluate. You have built or owned an eval suite for a non-deterministic system, and you can explain what you measured, how you produced ground truth, and what gated a release. Model and agent monitoring, drift detection. Cost and latency discipline: model tiering, caching, and the ability to say what a workload costs to run before it runs. Hands-on production experience with the Claude ecosystem —  Claude Code, CLAUDE.md, hooks, skills files.  Spec-driven development — writing the intent, constraints, and acceptance criteria before you let an agent build — is a strong plus. MCP: you can say why an agent would prefer it to a REST integration. Having authored a server is a plus. Nice to Have: Experience in one of the industries: financial services, insurance, healthcare. Consulting, professional services, or other embedded customer-facing delivery. AWS and Claude Code Certifications A2A: you can explain agent-to-agent interoperability CI/CD pipeline experience (GitHub Actions, GitLab CI) Practical experience with one or more use cases from the following: NLP, LLMs, and Recommendation engines. Experience in an additional language (Go, TypeScript, or Rust). Experience with Apache Spark, Apache Airflow, Kafkа Responsibilities: Work in a pair with an FDE and an FDX. Build and ship production GenAI systems into the customer’s environment (cloud-native data, LLM-based, and agentic AI solutions). Build and optimize RAG systems for production use cases Build the evaluation harness before you build the feature. Write production code across the stack — AI, backend services, data pipelines. We choose tools to fit the customer. Integrate AI components into backend services and RESTful APIs Take systems to production on AWS (GCP or Azure where the customer requires it): containerised, CI/CD. Implement LLMOps and AgentOps practices: agent tracing, prompt and version management, cost and latency monitoring, regression testing, drift detection Start from the blueprint, contribute to enablement and handover: clear documentation, runbooks, and pairing with the client engineers who will inherit the system. Feed reusable components and lessons back into the Provectus Blueprints Participate in technical discussions and architectural decisions Conduct model evaluation, improve failure modes you find, optimize model performance, efficiency, and reliability Mentor junior and mid-level AI engineers, conduct code reviews and share knowledge across the team through documentation, presentations, and workshops. What We Offer: The chance to shape how leading enterprises across LATAM, Europe, and North America adopt AI, from strategy through first deployment A forward-deployed model working in small, senior teams alongside FDE and FDX A growing AI delivery practice where you help build the tooling and frameworks, not just use them Remote-friendly culture Internal training programs with full support for Claude, AWS, and other professional certifications, conference attendance Career growth; we actively develop our engineers Access to the latest AI tools and premium subscriptions Long-term B2B collaboration Private medical insurance or a budget for your medical needs Paid sick leave, vacation, and public holidays Equipment and all the tech you need for comfortable, productive work How we hire: Intro conversation. The role, your background and aspirations, tech questions. Technical interview with live engineering sessions. Real problems, your own editor, you may use an LLM assistant HR Interview. Soft skills and expectations HM interview. Tech questions; a live engineering session is also possible

Location: New Jersey; New York; Pennsylvania; Connecticut; Massachusetts Workplace: remote About the role: At Provectus, you don’t prototype in a sandbox and hand a client a slide deck. You embed inside a bank, an insurer, or a health system, learn the work, and ship a Claude + AWS system into their real, regulated production — then you watch the business number move. We’re an Anthropic and AWS Partner with 150+ certified architects. That means the newest models, real budgets, and problems that matter — without the layers, politics, and years-to-impact of a hundred-thousand-person firm. 350+ builders, 100+ customers in production, founder-led since 2010, offices across North America, LATAM, and EMEA. Why Provectus Frontier access — Anthropic Preferred Partner: newest Claude models, real budgets, real problems. Real tasks (not scope), real deployment. You go where the operating problem is and ship production code to real users, not slide decks that sit in a drive. Embedded, not engaged. You are part of the customer’s team and inside their process — not a vendor running a project alongside it. Embedded is not staff-augmented. Autonomous. You own the work from framing through to production, visible to clients and founders.  You own the method; nobody hands you a ticket. Your experience stays portable. Every engagement is a new problem, often in a new industry, so your skills compound instead of narrowing to one product. No legacy-maintenance grind, not a single stack. Craft & growth — build beside senior FDEs on a Claude-certified team; a real platform (talks, OSS) to grow your name. We give you a stage: Claude certification, conference talks, open-source contributions, published work under your own name. What we look for: Typically 7+ years building software, a substantial share of it writing production code you were accountable for. You are hands-on today and intend to stay that way. Genuinely willing to spend time understanding someone else’s job before writing code; demonstrated ability to become conversant in an unfamiliar business function quickly. Shipped GenAI/LLM systems to production — not demos, not notebooks. You’ve handled the parts that get hard after the prototype works. You evaluate. You have built or owned an eval suite for a non-deterministic system, and you can explain what you measured and why. Strong engineering fundamentals — dropped into an unfamiliar codebase or language, you’re productive. Python and/or TypeScript proficiency; depth matters more than stack. Cloud-native delivery on AWS (GCP/Azure a plus), including the operational reality of a system someone else inherits. Pre-sales experience. Credible with senior stakeholders — you can hold a redesign conversation with a BU head and a scoping conversation with a CTO without losing either room. Comfort with ambiguity and ownership. Engagements start underspecified by design. Closing that gap is the job. Hands-on experience with Claude, Claude Code, and Cowork (skills, workflows, MCP, integrations, connectors, custom harnesses). Nice to have Containers and orchestration (Kubernetes/ECS) and CI/CD in production. IaC depth: AWS CDK, CloudFormation, Terraform. Open-source contributions or public writing on applied AI. What you’ll do: Spend the first weeks of each engagement in the operator’s seat, learning the constraints from the inside, then rebuild the function from first principles alongside the client and a Forward Deployed Executive — and you are the one who builds it. Be part of the customer’s team, inside their process; work on real tasks, not scoped deliverables, and own the method autonomously. You won’t start from zero. Provectus builds industry Blueprints — working systems that have already shipped for a customer in your industry. Your engagement starts from that baseline, and what you learn in the field goes back into it. Deliver across Cowork Activation, Agentic SDLC, and Blueprint Activations in Financial Services and Healthcare. Help shape how leading enterprises adopt AI, from strategy through first deployment; help build the tooling and frameworks of a growing AI delivery practice. What We Offer: Full-time. On-site visits to client’s facilities and our office in NYC 1/week Unlimited Vacation policy. Generous health, vision, and dental insurance. 401(K) matching plan. Base salary $180k–$200k + bonuses; OTE up to $250k. The salary range is determined through interviews and a review of the education, experience, knowledge, skills, and abilities of the applicant, and alignment with market data. We’re real: A person reads your application. We filter out AI-generated and fraudulent submissions automatically. Every application that clears that filter is read by someone on our team. We may ask for a quick ID verification. Fraudulent applications are common, and this protects real candidates. You’ll meet the people you’d work with. We host AI engineering events in New York and San Francisco. We'd be happy to talk in person. Your interviews are with Pavel Borobov, Solutions Architect, and Stepan Pushkarev, Co-Founder & CEO. How we hire: Recruitment interview. Tech interview with our FDE (30 min). A live engineering building session (1.5 hours). We hand you a business function and the constraints. You have to understand the job and rebuild it — then say what you’d build and how you’d know it worked. Use AI the way you normally would; we expect AI-generated code. What we look at is how you used AI, the judgment behind your choices, and what you researched beyond what any LLM already knows. 4. Soft skills interview. 5. Offer discussion.

Location: Serbia; Spain; Poland; North Macedonia; Armenia; Tbilisi Workplace: remote Provectus is a consulting and professional services firm with a focus on AI/ML based solutions. Provectus is an AWS Premier Partner and an Anthropic partner, helping enterprise clients move from AI experimentation to AI running in production. We are changing how we deliver. Instead of handing a client a design and waiting for someone else to build it, our engineers embed directly inside the client's team and ship the system themselves. We call this Forward Deployed delivery. Before you design anything, you spend real time with the people who do the work the system is meant to change, so the architecture reflects how the work actually happens, not how a requirements document described it. We are looking for a Senior Software Engineer or Solutions Architect with deep Python and Generative AI experience who wants to work this way: deep in the code, close to the client, and directly accountable for whether the system gets used. Direct client communication sits at the center of this role. Prior presales experience is a strong plus. What You Will Do: Embed with the client team for the length of the engagement, working inside their environment rather than alongside it from a distance Spend time with the people who do the job the system is meant to change before you design anything, so the architecture is grounded in how the work actually happens Write clean, production grade Python across AI integrations, backend services, and RESTful APIs (Flask, Django, or FastAPI) Design, build, and optimize RAG systems and agentic AI solutions running in production, not demos or notebooks Own system design and architectural decisions across the engagement: microservices vs monolith, sync vs event driven, SQL vs NoSQL Own the technical direction of the engagement from discovery through delivery Act as the client's direct technical point of contact: present architecture, defend trade offs, and push back on scope when it does not match timeline or budget Support presales activities where relevant, including discovery calls, technical proposals, scoping, and client facing demos Lead architecture reviews, produce technical design documents, and contribute to standards across the Python practice Feed what you learn back into Provectus's blueprint library, so the next engagement starts further ahead Mentor engineers, lead code reviews, and share knowledge across the team What You Will Bring: 7+ years building and running production systems; hands on production experience is required, demo only or POC only backgrounds will not be considered Production experience with RAG systems and LLM based agentic workflows that are live and running, not prototypes Backend development experience with Flask, Django REST, or FastAPI Recent, hands on experience with AWS (SageMaker, Bedrock, Lambda, ECS, S3, SQS, ECR, or similar); GCP considered Experience with LLM APIs (OpenAI, Anthropic Claude, or AWS Bedrock) Strong Python proficiency: OOP, design patterns, clean architecture, and performance optimization Demonstrated experience making and defending system design and architectural trade off decisions Comfortable owning direct communication with client stakeholders, without a project manager relaying the conversation Genuine willingness to spend time close to the people who do the job before you design a system to change it, and comfort with an engagement that starts underspecified by design Strong testing practices: pytest, mocking, and integration tests for AI systems Experience with Docker and Kubernetes Understanding of LLM evaluation techniques and quality assurance approaches Experience deploying and maintaining AI/ML models in production environments Already using AI assisted development tools in your daily workflow (Claude Code, Copilot, or similar) Proactive and self directed; you own outcomes end to end and spot problems before they are handed to you B2+ English, comfortable collaborating across distributed, multicultural teams Nice to Have: Exposure to Financial Services or Healthcare and Life Sciences, our two primary practice verticals Presales experience: cost estimation, cloud architecture cost optimization, scoped and phased delivery plans Prior consulting, professional services, or other embedded client facing delivery Experience with React or Vue AWS or Claude Code certifications Experience with Streamlit or Gradio for AI prototyping Modern Python tooling (ruff, uv, pyproject.toml, pyright) CI/CD pipeline experience (GitHub Actions, GitLab CI) Experience in an additional language (Go, Node.js, or Rust) What We Offer: Remote-friendly culture Internal training programs with full support for Claude, AWS, and other professional certifications, conference attendance Career growth; we actively develop our engineers Access to the latest AI tools and premium subscriptions Long-term B2B collaboration Private medical insurance or a budget for your medical needs Paid sick leave, vacation, and public holidays Equipment and all the tech you need for comfortable, productive work

Location: New York, New York; New York City Metropolitan Area; New Jersey; Connecticut; San Francisco Bay Area, California Workplace: remote About the role: At Provectus, you don’t prototype in a sandbox and hand a client a slide deck. You embed inside a bank, an insurer, or a health system, learn the work, and ship a Claude + AWS system into their real, regulated production — then you watch the business number move. We’re an Anthropic and AWS Partner with 150+ certified architects. That means the newest models, real budgets, and problems that matter — without the layers, politics, and years-to-impact of a hundred-thousand-person firm. 350+ builders, 100+ customers in production, founder-led since 2010, offices across North America, LATAM, and EMEA. Why Provectus Frontier access — Anthropic Preferred Partner: newest Claude models, real budgets, real problems Real tasks (not scope), real deployment. You go where the operating problem is and ship production code to real users, not slide decks that sit in a drive Embedded, not engaged. You are part of the customer’s team and inside their process — not a vendor running a project alongside it. Embedded is not staff-augmented Autonomous. You own the work from framing through to production, visible to clients and founders.  You own the method; nobody hands you a ticket Your experience stays portable. Every engagement is a new problem, often in a new industry, so your skills compound instead of narrowing to one product. No legacy-maintenance grind, not a single stack Craft & growth — build beside senior FDEs on a Claude-certified team; a real platform (talks, OSS) to grow your name We give you a stage: Claude certification, conference talks, open-source contributions, published work under your own name What You’ll Do: Spend the first weeks of each engagement in the operator’s seat, learning the constraints from the inside, then rebuild the function from first principles alongside the client and a Forward Deployed Executive — and you are the one who builds it Be part of the customer’s team, inside their process; work on real tasks, not scoped deliverables, and own the method autonomously You won’t start from zero. Provectus builds industry Blueprints — working systems that have already shipped for a customer in your industry. Your engagement starts from that baseline, and what you learn in the field goes back into it Deliver across Cowork Activation, Agentic SDLC, and Blueprint Activations in Financial Services and Healthcare Help shape how leading enterprises adopt AI, from strategy through first deployment; help build the tooling and frameworks of a growing AI delivery practice What We Look For: Typically 7+ years building software, a substantial share of it writing production code you were accountable for. You are hands-on today and intend to stay that way Genuinely willing to spend time understanding someone else’s job before writing code; demonstrated ability to become conversant in an unfamiliar business function quickly Shipped GenAI/LLM systems to production — not demos, not notebooks. You’ve handled the parts that get hard after the prototype works You evaluate. You have built or owned an eval suite for a non-deterministic system, and you can explain what you measured and why Strong engineering fundamentals — dropped into an unfamiliar codebase or language, you’re productive. Python and/or TypeScript proficiency; depth matters more than stack Cloud-native delivery on AWS (GCP/Azure a plus), including the operational reality of a system someone else inherits Pre-sales experience. Credible with senior stakeholders — you can hold a redesign conversation with a BU head and a scoping conversation with a CTO without losing either room Comfort with ambiguity and ownership. Engagements start underspecified by design. Closing that gap is the job Hands-on experience with Claude, Claude Code, and Cowork (skills, workflows, MCP, integrations, connectors, custom harnesses) Nice to have: Containers and orchestration (Kubernetes/ECS) and CI/CD in production IaC depth: AWS CDK, CloudFormation, Terraform Open-source contributions or public writing on applied AI What We Offer: Full-time employment On-site visits to clients’ facilities and our office in NYC 1/week Unlimited Vacation policy Generous health, vision, and dental insurance 401(K) matching plan Base salary $180k–$200k + bonuses; OTE up to $250k. The salary range is determined through interviews and a review of the education, experience, knowledge, skills, and abilities of the applicant, and alignment with market data No Visa sponsorship How We Hire: Recruitment interview. Tech interview with our FDE (30 min). A live engineering building session (1.5 hours). We hand you a business function and the constraints. You have to understand the job and rebuild it — then say what you’d build and how you’d know it worked. Use AI the way you normally would; we expect AI-generated code. What we look at is how you used AI, the judgment behind your choices, and what you researched beyond what any LLM already knows. 4. Soft skills interview. 5. Offer discussion. We’re real: A person reads your application. We filter out AI-generated and fraudulent submissions automatically. Every application that clears that filter is read by someone on our team We may ask for a quick ID verification. Fraudulent applications are common, and this protects real candidates You’ll meet the people you’d work with. We host AI engineering events in New York and San Francisco. We'd be happy to talk in person. Your interviews are with Pavel Borobov, Solutions Architect, and Stepan Pushkarev, Co-Founder & CEO

Location: New York; Massachusetts; Washington DC-Baltimore Area; Pennsylvania Workplace: remote About the role: We’re an Anthropic and AWS Partner with 150+ certified architects. That means the newest models, real budgets, and problems that matter — without the layers, politics, and years-to-impact of a hundred-thousand-person firm. 350+ builders, 100+ customers in production, founder-led since 2010, offices across North America, LATAM, and EMEA. Why Provectus Frontier access — Anthropic Preferred Partner: newest Claude models, real budgets, real problems Real tasks (not scope), real deployment. You go where the operating problem is and ship production code to real users, not slide decks that sit in a drive Embedded, not engaged. You are part of the customer’s team and inside their process — not a vendor running a project alongside it. Embedded is not staff-augmented Autonomous. You own the work from framing through to production, visible to clients and founders.  You own the method; nobody hands you a ticket Your experience stays portable. Every engagement is a new problem, often in a new industry, so your skills compound instead of narrowing to one product. No legacy-maintenance grind, not a single stack Craft & growth — build beside senior FDEs on a Claude-certified team; a real platform (talks, OSS) to grow your name We give you a stage: Claude certification, conference talks, open-source contributions, published work under your own name What You’ll Do: Spend the first weeks of each engagement in the operator’s seat, learning the constraints from the inside, then rebuild the function from first principles alongside the client and a Forward Deployed Executive — and you are the one who builds it Be part of the customer’s team, inside their process; work on real tasks, not scoped deliverables, and own the method autonomously You won’t start from zero. Provectus builds industry Blueprints — working systems that have already shipped for a customer in your industry. Your engagement starts from that baseline, and what you learn in the field goes back into it Deliver across Cowork Activation, Agentic SDLC, and Blueprint Activations in Financial Services and Healthcare Help shape how leading enterprises adopt AI, from strategy through first deployment; help build the tooling and frameworks of a growing AI delivery practice What We Look For: Typically 7+ years building software, a substantial share of it writing production code you were accountable for. You are hands-on today and intend to stay that way Genuinely willing to spend time understanding someone else’s job before writing code; demonstrated ability to become conversant in an unfamiliar business function quickly Shipped GenAI/LLM systems to production — not demos, not notebooks. You’ve handled the parts that get hard after the prototype works You evaluate. You have built or owned an eval suite for a non-deterministic system, and you can explain what you measured and why Strong engineering fundamentals — dropped into an unfamiliar codebase or language, you’re productive. Python and/or TypeScript proficiency; depth matters more than stack Cloud-native delivery on AWS (GCP/Azure a plus), including the operational reality of a system someone else inherits Pre-sales experience. Credible with senior stakeholders — you can hold a redesign conversation with a BU head and a scoping conversation with a CTO without losing either room Comfort with ambiguity and ownership. Engagements start underspecified by design. Closing that gap is the job Hands-on experience with Claude, Claude Code, and Cowork (skills, workflows, MCP, integrations, connectors, custom harnesses) Nice to have: Containers and orchestration (Kubernetes/ECS) and CI/CD in production IaC depth: AWS CDK, CloudFormation, Terraform Open-source contributions or public writing on applied AI What We Offer: Full-time employment On-site visits to clients’ facilities and our office in NYC 1/week Unlimited Vacation policy Generous health, vision, and dental insurance 401(K) matching plan Base salary $180k–$200k + bonuses; OTE up to $250k. The salary range is determined through interviews and a review of the education, experience, knowledge, skills, and abilities of the applicant, and alignment with market data No Visa sponsorship How we hire: Recruitment interview. Tech interview with our FDE (30 min). A live engineering building session (1.5 hours). We hand you a business function and the constraints. You have to understand the job and rebuild it — then say what you’d build and how you’d know it worked. Use AI the way you normally would; we expect AI-generated code. What we look at is how you used AI, the judgment behind your choices, and what you researched beyond what any LLM already knows. 4. Soft skills interview. 5. Offer discussion. We’re real: A person reads your application. We filter out AI-generated and fraudulent submissions automatically. Every application that clears that filter is read by someone on our team We may ask for a quick ID verification. Fraudulent applications are common, and this protects real candidates You’ll meet the people you’d work with. We host AI engineering events in New York and San Francisco. We'd be happy to talk in person. Your interviews are with Pavel Borobov, Solutions Architect, and Stepan Pushkarev, Co-Founder & CEO

Location: Москва Это возможность возглавить исследовательское направление в команде, занимающейся проектами в новом, динамично развивающемся направлении безопасности искусственного интеллекта. Ваши задачи будут связаны с поиском и исследованием уязвимостей GenAI и AI-приложений с их использованием, в том числе AI-агентов. В коллективе опытных экспертов вы будете оценивать безопасность разрабатываемого ПО, предлагать инструменты для автоматизации исследования и тестирования безопасности GenAI-моделей, а также формировать и развивать исследовательскую команду, участвуя в международных и внутренних CTF. Это не административная роль, мы ищем «играющего тренера», который сам может написать первый прототип, проверить гипотезу об угрозе и решить таск на CTF - и только потом делегирует масштабирование команде. Если ваш опыт ограничивается управлением процессами без hands-on разработки и исследований, эта позиция не подойдёт. Обязанности ### Вам предстоит: * исследование угроз GenAI: создание, поддержка и развитие инструментов для анализа и поиска угроз, в т.ч.: \-разработка средств автоматизации сбора, хранения и анализа новых публикаций, инцидентов и методов атак на GenAI (краулинг, парсинг информации, разработка и управление системами хранения) \-создание и тестирование прототипов (Proof of Concept) для поиска уязвимостей GenAI (скрипты, основные библиотеки GenAI, AI-агенты для генерации кода) * DevOps: CI/CD, контроль версий (git), тестирование качества, инфраструктура для работы (linux) * внедрение практик AI PDLC: исследование и применение методов и инструментов создания и развития программного обеспечения с применением передовых ИИ-моделей * погружение в экосистему: изучение и практическое применение передовых фреймворков и библиотек для работы с LLM и AI-агентами * управление командой: формирование и развитие исследовательской команды (3–5 человек), приоритизация исследований, защита решений перед руководством и смежными командами * CTF и хакатоны: формирование и капитанство команды на международных CTF по AI/ML security, трансляция найденных техник в работу направления * публикации и нормативные документы: оформление результатов исследований, вклад в нормативные документы безопасности AI. Требования ### Что для нас важно: * профиль и образование: законченное высшее образование в области информационных технологий, компьютерных наук, прикладной математики, программной инженерии или смежных направлений * продвинутые навыки программирования: понимание принципов проектирования и разработки программного обеспечения, опыт в написании и чтении программного кода (понятный, поддерживаемый) * навыки DevOps: работа с git, администрирование linux, автоматизация этапов жизненного цикла разработки ПО * LLM и AI-агенты: экспертный уровень применения LLM для разработки ПО, глубокое понимание архитектуры агентов и их векторов атак * английский язык на уровне не менее чтения технической документации, статей и участия в международных CTF * опыт командной работы в проектах разработки: канбан (спринты, трекинг задач), jira/git/confluence, презентации и умение понятно объяснять свои идеи команде/руководству и согласовывать требования * Adversarial ML и red teaming LLM: практический опыт поиска уязвимостей в GenAI-моделях (prompt injection, jailbreaking, tool poisoning, data extraction) * знание инструментария для LLM security: Giskard, Garak, LLM Guard и др. * CTF и хакатоны: участие в международных CTF по AI/ML security (желательно — победы или призовые места). Условия ### Мы предлагаем: * комфортный современный офис ул. Вавилова 19 * офисный формат работы * ежегодный пересмотр зарплаты, годовая премия * корпоративный спортзал и зоны отдыха * более 400 образовательных программ СберУниверситета для профессионального и карьерного развития * программа адаптации и помощь руководителя на старте * расширенный ДМС, льготное страхование для семьи и корпоративная пенсионная программа * гибкий дисконт по ипотечному кредиту, равный 1/3 ключевой ставки ЦБ * подписка Прайм с возможностью совместного использования на трёх близких * вознаграждение за рекомендацию друзей в команду Сбера.

Location: Spain; Serbia; Poland; Ukraine; Portugal; North Macedonia; Bosnia and Herzegovina; Romania Workplace: remote About Provectus Provectus is an AWS Premier Partner and an Anthropic Strategic Partner, working at the frontier of applied AI. We help enterprises turn Claude, agentic systems, and their own data into measurable business outcomes — through bespoke applications, managed services, and advisory engagements. With offices in North America, LATAM, and EMEA, we partner with clients worldwide.   Our work centers on two verticals — Financial Services & Insurance and Healthcare & Life Sciences — where we deploy five pre-built AI Blueprints: Submission Flow, Portfolio Lens, Asset Flow, Revenue Flow, and Evidence Lens. Each Blueprint rebuilds a critical business process front to back, shipped from working code and tuned to a client's specific book, regulators, and operating posture.   Our team holds 100+ AWS certifications, is Claude Code certified, and co-delivers Anthropic's Agentic SDLC program, Cowork Activation, and AI Blueprint engagements. Where this role sits You will work in a small, senior pod alongside an FDE and an FDX, Forward Deployed Executives (FDX) own the commercial relationship and the business outcome. Works alongside the client's leadership or C-suite level to move the client's KPIs. Forward Deployed Engineers (FDE) embed with a client, map the client workflow, identify the business problem underneath it, design and build a working AI solution, present to the client, and transfer the knowledge to the client's team. Owns technical direction of the whole solution. Senior AI Engineer. When an FDE comes back from the client with the business problem, you will help turn that into an agentic system that runs in production and will be responsible for evaluation, observability, and guardrails.  You'll have real ownership of components and of the technical decisions inside them. Requirements: Mindset Proactive and self-directed; you push for clarity rather than waiting for a ticket Excellent communication and problem-solving skills Comfort with ambiguity and ownership. B2+ English, comfortable collaborating across distributed, multicultural teams. Technical depth 5+ years in software or ML engineering, with production systems you were accountable for . Solid AI/ML foundations. You understand what the models do well enough to reason about failure modes. Shipped to production LLM applications and agentic workflows — not demos, not POCs, not notebooks. Agentic orchestration: multi-step workflows, graph-based orchestration, tool use, state management, and recovery from partial failure Experience with LLM APIs (Anthropic, AWS Bedrock, or OpenAI) and agent frameworks. Experience building and optimizing RAG systems in production. Strong engineering fundamentals. Full-stack mindset, comfortable across AI, backend development, and cloud infrastructure. Python and/or TypeScript proficiency; depth matters more than stack. Dropped into an unfamiliar codebase, you're productive. Hands-on AWS in production: Bedrock, Bedrock AgentCore, Lambda, ECS, S3, SQS, ECR, or similar. GCP or Azure is a plus. Cloud-native delivery: containers, ECS or Kubernetes, IaC, and CI/CD applied to AI pipelines. You evaluate. You have built or owned an eval suite for a non-deterministic system, and you can explain what you measured, how you produced ground truth, and what gated a release. Model and agent monitoring, drift detection. Cost and latency discipline: model tiering, caching, and the ability to say what a workload costs to run before it runs. Hands-on production experience with the Claude ecosystem —  Claude Code, CLAUDE.md, hooks, skills files.  Spec-driven development — writing the intent, constraints, and acceptance criteria before you let an agent build — is a strong plus. MCP: you can say why an agent would prefer it to a REST integration. Having authored a server is a plus. Responsibilities: Work in a pair with an FDE and an FDX. Build and ship production GenAI systems into the customer’s environment (cloud-native data, LLM-based, and agentic AI solutions). Build and optimize RAG systems for production use cases Build the evaluation harness before you build the feature. Write production code across the stack — AI, backend services, data pipelines. We choose tools to fit the customer. Integrate AI components into backend services and RESTful APIs Take systems to production on AWS (GCP or Azure where the customer requires it): containerised, CI/CD. Implement LLMOps and AgentOps practices: agent tracing, prompt and version management, cost and latency monitoring, regression testing, drift detection Start from the blueprint, contribute to enablement and handover: clear documentation, runbooks, and pairing with the client engineers who will inherit the system. Feed reusable components and lessons back into the Provectus Blueprints Participate in technical discussions and architectural decisions Conduct model evaluation, improve failure modes you find, optimize model performance, efficiency, and reliability Mentor junior and mid-level AI engineers, conduct code reviews and share knowledge across the team through documentation, presentations, and workshops. Nice to Have: Experience in one of the industries: financial services, insurance, healthcare. Consulting, professional services, or other embedded customer-facing delivery. AWS and Claude Code Certifications A2A: you can explain agent-to-agent interoperability CI/CD pipeline experience (GitHub Actions, GitLab CI) Practical experience with one or more use cases from the following: NLP, LLMs, and Recommendation engines. Experience in an additional language (Go, TypeScript, or Rust). Experience with Apache Spark, Apache Airflow, Kafkа What We Offer: The chance to shape how leading enterprises across LATAM, Europe, and North America adopt AI, from strategy through first deployment A forward-deployed model working in small, senior teams alongside FDE and FDX A growing AI delivery practice where you help build the tooling and frameworks, not just use them Remote-friendly culture Internal training programs with full support for Claude, AWS, and other professional certifications, conference attendance Career growth; we actively develop our engineers Access to the latest AI tools and premium subscriptions Long-term B2B collaboration Private medical insurance or a budget for your medical needs Paid sick leave, vacation, and public holidays Equipment and all the tech you need for comfortable, productive work How we hire:   Intro conversation. The role, your background and aspirations, tech questions. Technical interview with live engineering sessions. Real problems, your own editor, you may use an LLM assistant  HR Interview. Soft skills and expectations HM interview. Tech questions; a live engine

Location: Armenia; Spain; Serbia; Poland; Kyiv City; Kyiv; Odesa; North Macedonia Workplace: remote About Provectus Provectus is an AWS Premier Partner and an Anthropic Strategic Partner, working at the frontier of applied AI. We help enterprises turn Claude, agentic systems, and their own data into measurable business outcomes — through bespoke applications, managed services, and advisory engagements. With offices in North America, LATAM, and EMEA, we partner with clients worldwide.   Our work centers on two verticals — Financial Services & Insurance and Healthcare & Life Sciences — where we deploy five pre-built AI Blueprints: Submission Flow, Portfolio Lens, Asset Flow, Revenue Flow, and Evidence Lens. Each Blueprint rebuilds a critical business process front to back, shipped from working code and tuned to a client's specific book, regulators, and operating posture.   Our team holds 100+ AWS certifications, is Claude Code certified, and co-delivers Anthropic's Agentic SDLC program, Cowork Activation, and AI Blueprint engagements. Where this role sits You will work in a small, senior pod alongside an FDE and an FDX, Forward Deployed Executives (FDX) own the commercial relationship and the business outcome. Works alongside the client's leadership or C-suite level to move the client's KPIs. Forward Deployed Engineers (FDE) embed with a client, map the client workflow, identify the business problem underneath it, design and build a working AI solution, present to the client, and transfer the knowledge to the client's team. Owns technical direction of the whole solution. Senior AI Engineer. When an FDE comes back from the client with the business problem, you will help turn that into an agentic system that runs in production and will be responsible for evaluation, observability, and guardrails.  You'll have real ownership of components and of the technical decisions inside them. Requirements: Mindset Proactive and self-directed; you push for clarity rather than waiting for a ticket Excellent communication and problem-solving skills Comfort with ambiguity and ownership. B2+ English, comfortable collaborating across distributed, multicultural teams. Technical depth 5+ years in software or ML engineering, with production systems you were accountable for . Solid AI/ML foundations. You understand what the models do well enough to reason about failure modes. Shipped to production LLM applications and agentic workflows — not demos, not POCs, not notebooks. Agentic orchestration: multi-step workflows, graph-based orchestration, tool use, state management, and recovery from partial failure Experience with LLM APIs (Anthropic, AWS Bedrock, or OpenAI) and agent frameworks. Experience building and optimizing RAG systems in production. Strong engineering fundamentals. Full-stack mindset, comfortable across AI, backend development, and cloud infrastructure. Python and/or TypeScript proficiency; depth matters more than stack. Dropped into an unfamiliar codebase, you're productive. Hands-on AWS in production: Bedrock, Bedrock AgentCore, Lambda, ECS, S3, SQS, ECR, or similar. GCP or Azure is a plus. Cloud-native delivery: containers, ECS or Kubernetes, IaC, and CI/CD applied to AI pipelines. You evaluate. You have built or owned an eval suite for a non-deterministic system, and you can explain what you measured, how you produced ground truth, and what gated a release. Model and agent monitoring, drift detection. Cost and latency discipline: model tiering, caching, and the ability to say what a workload costs to run before it runs. Hands-on production experience with the Claude ecosystem —  Claude Code, CLAUDE.md, hooks, skills files.  Spec-driven development — writing the intent, constraints, and acceptance criteria before you let an agent build — is a strong plus. MCP: you can say why an agent would prefer it to a REST integration. Having authored a server is a plus. Responsibilities: Work in a pair with an FDE and an FDX. Build and ship production GenAI systems into the customer’s environment (cloud-native data, LLM-based, and agentic AI solutions). Build and optimize RAG systems for production use cases Build the evaluation harness before you build the feature. Write production code across the stack — AI, backend services, data pipelines. We choose tools to fit the customer. Integrate AI components into backend services and RESTful APIs Take systems to production on AWS (GCP or Azure where the customer requires it): containerised, CI/CD. Implement LLMOps and AgentOps practices: agent tracing, prompt and version management, cost and latency monitoring, regression testing, drift detection Start from the blueprint, contribute to enablement and handover: clear documentation, runbooks, and pairing with the client engineers who will inherit the system. Feed reusable components and lessons back into the Provectus Blueprints Participate in technical discussions and architectural decisions Conduct model evaluation, improve failure modes you find, optimize model performance, efficiency, and reliability Mentor junior and mid-level AI engineers, conduct code reviews and share knowledge across the team through documentation, presentations, and workshops. Nice to Have: Experience in one of the industries: financial services, insurance, healthcare. Consulting, professional services, or other embedded customer-facing delivery. AWS and Claude Code Certifications A2A: you can explain agent-to-agent interoperability CI/CD pipeline experience (GitHub Actions, GitLab CI) Practical experience with one or more use cases from the following: NLP, LLMs, and Recommendation engines. Experience in an additional language (Go, TypeScript, or Rust). Experience with Apache Spark, Apache Airflow, Kafkа What We Offer: The chance to shape how leading enterprises across LATAM, Europe, and North America adopt AI, from strategy through first deployment A forward-deployed model working in small, senior teams alongside FDE and FDX A growing AI delivery practice where you help build the tooling and frameworks, not just use them Remote-friendly culture Internal training programs with full support for Claude, AWS, and other professional certifications, conference attendance Career growth; we actively develop our engineers Access to the latest AI tools and premium subscriptions Long-term B2B collaboration Private medical insurance or a budget for your medical needs Paid sick leave, vacation, and public holidays Equipment and all the tech you need for comfortable, productive work How we hire:   Intro conversation. The role, your background and aspirations, tech questions. Technical interview with live engineering sessions. Real problems, your own editor, you may use an LLM assistant  HR Interview. Soft skills and expectations HM interview. Tech questions; a live engine