Вакансии

Location: Athens Workplace: Remote About the company: KIEFER is building Greece's integrated AI ecosystem. From renewable energy infrastructure and AI systems to robotics and enterprise applications, we connect the technologies that power Europe's intelligent future. Founded in 2014, KIEFER has delivered 600MW+ of energy projects and is now developing sovereign AI infrastructure, enterprise AI products and physical AI systems for Greece and Southeast Europe. About the role We are looking for an ML Engineer to improve the agent and long-context LLM capabilities behind Kiefer Tech’s AI products. You will work hands-on across post-training, data, evaluation, agent design, and inference, with the goal of making multi-step tasks more reliable in production. What you will do - Run post-training experiments using supervised fine-tuning, preference optimization, reinforcement learning, or other suitable methods. - Improve how agents plan, use tools, manage execution, and recover from errors. - Build benchmarks, verifiers, and diagnostic tools to measure agent performance and identify failures. - Generate and curate training data, including agent trajectories and difficult cases from target workflows. - Evaluate and improve long-context behavior, memory and retrieval interactions, and reliability over multi-step tasks. - Work with ML, Product, and Engineering to improve inference and bring successful experiments into production with clear regression checks. What you will need - 3+ years of relevant ML or research engineering experience, with clear technical ownership and measurable results. - Hands-on LLM post-training experience in at least one area: supervised fine-tuning, preference optimization, reinforcement learning, or a comparable method. - Experience building agent systems involving tool use, multi-step execution, and error recovery. - Strong skills in model and agent evaluation, failure analysis, and the creation or curation of training data. - Strong Python, hands-on experience with PyTorch, JAX, TensorFlow, or a similar framework, and solid software engineering skills. - Experience improving or deploying model-backed systems, with an understanding of LLM inference performance and cost. Nice to have - Reward modeling or verifier-based training. - Long-context training, memory systems, retrieval, or context compression. - Distributed training or high-performance LLM serving. - Experience delivering agents for complex enterprise, coding, or research workflows. What is there for you - Compensation: Competitive package aligned with talent benchmarks. - Impact: Hands-on work shaping AI capabilities for Kiefer Tech’s products and real-world applications. - AI-native environment: Real challenges across model development, evaluation, and production AI systems. - Culture: Engineering-first, high autonomy, low bureaucracy, and space to shape meaningful AI products.

Location: Remote - United States Reddit is a community of communities. It’s built on shared interests, passion, and trust, and is home to the most open and authentic conversations on the internet. Every day, Reddit users submit, vote, and comment on the topics they care most about. With 100,000+ active communities and approximately 130 million daily active unique visitors, Reddit is one of the internet’s largest sources of information. For more information, visit www.redditinc.com . Location: Reddit has a flexible first workforce. Don't live near our office? No worries: you can work remotely from anywhere in the US or Canada. About the Team Reddit's Ads ML Efficiency team is focused on making machine learning workloads faster, more scalable, and more cost-efficient, enabling our Ads ML teams to develop, launch, and scale increasingly sophisticated models. This role sits at the intersection of ML workloads, execution systems, and engineering leverage. The engineer will partner closely with ranking teams, serving owners, and ML Platform to identify important bottlenecks, implement targeted improvements, and connect execution and systems improvements to measurable end-to-end outcomes. What you’ll do Own end-to-end performance optimization initiatives for high-impact production ML workloads, from bottleneck identification and implementation through validation and production rollout. Profile and diagnose ML performance bottlenecks across model execution, frameworks, runtimes, memory, CPU/GPU utilization, and infrastructure using rigorous benchmarking and measurement. Design and implement performance improvements involving inference execution, runtime behavior, batching, memory management, precision, compiler integration, custom operators, kernels, or other relevant techniques. Improve ML training and serving efficiency and scalability, optimizing latency, throughput, hardware utilization, and cost while maintaining model quality and production reliability. Support faster and safer model launches through performance benchmarking, load testing, regression detection, and launch-readiness improvements. Partner with model engineers and platform teams to identify high-leverage opportunities, integrate optimizations safely, and validate measurable end-to-end impact. Build reusable performance capabilities, including benchmarks, tooling, optimization patterns, and engineering guidance that accelerate future model development and deployment. Contribute to technical direction and mentor engineers through hands-on collaboration, code reviews, performance debugging, and knowledge sharing. What we’re looking for Hands-on ML performance engineering experience: Demonstrated ability to diagnose and implement meaningful performance improvements in production or production-relevant ML workloads. Deep expertise in at least one ML performance area, such as inference execution, ML runtimes, compiler integration, GPU/kernel optimization, memory systems, or training execution. Strong profiling and benchmarking skills: Ability to identify bottlenecks systematically, establish reliable baselines, and validate measurable end-to-end improvements. Strong software engineering fundamentals: Experience implementing, debugging, testing, and maintaining performance-sensitive systems, with proficiency in Python and relevant systems programming technologies. ML systems understanding: Familiarity with production ML frameworks and the performance trade-offs across model execution, runtime, memory, hardware, and infrastructure. Production engineering judgment: Ability to balance performance improvements with numerical correctness, model quality, reliability, maintainability, and safe rollout. Good customer and platform instincts: can solve concrete bottlenecks while keeping maintainability, adoption, and future reuse in mind. Strong communication: able to explain tradeoffs clearly to engineers and partner teams. Nice-to-have Experience optimizing large-scale inference or training workloads, particularly in recommendation systems, ranking, search, or retrieval. Experience with PyTorch, including performance optimization using technologies such as Dynamo, Inductor, AOTI, torch.export, or custom operators. Experience with inference runtimes, GPU execution, or performance technologies such as TensorRT, ONNX Runtime, CUDA, Triton, LLVM/MLIR, or equivalent systems. Experience with hardware-aware optimization, including batching, memory management, operator fusion, mixed precision, quantization, or accelerator efficiency. Experience with distributed-training performance, including GPU utilization, communication, data loading, memory efficiency, or training step-time optimization. Experience building performance benchmarks, profiling tools, regression detection, launch-readiness capabilities, or cost observability toolings for ML workloads. Experience working across ML modeling, serving, and platform infrastructure teams. Specialized compiler internals or custom CUDA/Triton kernel development are valuable but not required. We value demonstrated performance engineering depth, sound technical judgment, and measurable workload impact over familiarity with any particular tool or framework. Benefits: Global Benefit programs that fit your lifestyle, from workspace to professional development to caregiving support Family Planning Support Gender-Affirming Care Mental Health & Coaching Benefits Group Personal Pension Scheme with Employer match Private Medical and Dental Scheme Income Replacement Programs Bike to Work scheme Flexible Vacation & Paid Volunteer Time Off Generous Paid Parental Leave Pay Transparency: This job posting may span more than one career level. In addition to base salary, this job is eligible to receive equity in the form of restricted stock units, and depending on the position offered, it may also be eligible to receive a commission. Additionally, Reddit offers a wide range of benefits to U.S.-based employees, including medical, dental, and vision insurance, 401(k) program with employer match, generous time off for vacation, and parental leave. To learn more, please visit https://www.redditinc.com/careers/ . To provide greater transparency to candidates, we share base salary ranges for all US-based job postings regardless of state. We set standard base pay ranges for all roles based on function, level, and country location, benchmarked against similar stage growth companies. Final offer amounts are determined by multiple factors including, skills, depth of work experience and relevant licenses/credentials, and may vary from the amounts listed below. The base salary range for this position is: $216,700 — $303,400 USD In select roles and locations, the interviews will be recorded, transcribed and summarized by artificial intelligence (AI). You will have the opportunity to opt out of recording, transcription and summarization prior to any scheduled interviews. During the interview, we will collect the following categories of personal information: Identifiers, Professional and Employment-Related Information, Sensory Information (audio/video recording), and any other categories of personal information you choose to share with us. We will use this information to evaluate your application for employment or an independent contractor role, as applicable.  We will not sell your personal information or disclose it to any third party for their marketing purposes.  We will delete any recording of your interview promptly after making a hiring decision.  For more information about how we will handle your personal information, including our retention of it, please refer to our Candidate Privacy Policy for Potential Employees and Contractors . Reddit is proud to be an equal opportunity employer, and is committed to building a workforce representative of the diverse communities we serve.  Reddit is committed to providing reasonable accommodations for qualified individuals with disabilities and disabled veterans in our job application procedures. If, due to a disability, you need an accommodation during the interview process, please let your recruiter know.

Location: Serbia; Spain Social Discovery Group (SDG) is a group of social discovery companies. SDG solves the problems of loneliness, isolation, and disconnection - transforming virtual intimacy into the new normal. SDG’s products redefine the way people interact and connect with one another. Our portfolio includes social entertainment platforms designed to connect people online across different cultures and regions of the world. We bring together a team of like-minded people and IT professionals who specialize in creating and developing globally impactful social discovery products. Our international team of digital nomads works remotely from all over the world. We’re proud to be a two-time “Great Place to Work” winner (USA & Japan, 2024–2025) and a Top-5 Company for Work-From-Anywhere Jobs (FlexJobs, 2025). We are looking for Machine Learning Engineer. Your main tasks will be: Set the ML direction: choose which strategic bets to make and tie each one to clear business needs. Drive LLM post-training and model adaptation end to end, from dataset construction to fine-tuning and preference optimization. Design and build agent harnesses and tool-using LLM systems. Shape multi-agent development workflows that make the team faster and more effective. Build evaluation frameworks and LLM-as-a-judge systems that reliably predict real-world performance. Improve dialogue quality across memory, emotional intelligence, persona consistency, safety and anti-loop behavior. Ship production systems that directly shape user experience, retention and monetization. We expect from you: Python expert level: clean code, typing, async, multiprocessing, profiling, debugging, notebooks + production code.Trainsformers. Training and fine-tuning LLMs. SFT, LoRA/QLoRA, PEFT. Preference optimization: DPO / ORPO / SimPO / CPO / GRPO or willingness to quickly go deep. Building datasets for LLM training: cleaning, filtering, deduplication, preference pairs, synthetic data, hard negative mining. LLM evaluation: offline evals, pairwise evals, LLM judges, human annotation. Agent harness design: tool calling, function calling, structured outputs, retries, routing, state management, memory/context, guardrails. Production LLM integration: APIs, streaming, batching, fallbacks, cost/latency trade-offs. Observability for agents: traces, tool-call logs, transcript analysis, failure clustering. Strong understanding of transformer architecture and modern LLM behavior. Accuracy with dtype/device/batching: fp32/fp16/bf16, quantization, CUDA memory, avoiding dtype mismatch bugs. Git, Docker, Linux, basic MLOps. Fluent Russian, ready to work in CET (±2) hours. Nice to have: experience beyond text (computer vision, image generation, multimodal), long-running conversations and character consistency, vLLM / TGI / SGLang, DeepSpeed / FSDP / Accelerate, quantization, safety classifiers. What do we offer: REMOTE OPPORTUNITY to work full-time; The initial pay level or pay range for this role will be shared with candidates during the recruitment process and before the commencement of employment; Vacation 28 calendar days per year; 7 wellness days per year (time off) that can be used to deal with household issues, to lie down and recover without taking sick leave; Bonuses up to $5000 for recommending successful applicants for positions in the company; 50% payment for professional training, international conferences, and meetings; Corporate discount for English lessons; ​ Health benefits. According to the paychecks, if you are not eligible for corporate medical insurance, the company will compensate you with up to $ 1,000 gross per year per employee. This can be spent on self-purchase of health insurance or on doctor’s fees for yourself and close relatives (spouse, children); ​ Workplace organization. The company provides all employees with an equipped workplace and all the necessary equipment (table, armchair, wifi, etc.) in our offices or co-working locations. In the other locations, the company provides reimbursement of workplace costs up to $ 1000 gross once every 3 years, according to the paychecks. This money can be spent on the rent of the co-working room, on equipping the working place at home (desk, chair, Internet, etc.) during those 3 years; Internal gamified gratitude system: receive bonuses from colleagues and exchange them for our merchandise, team building activities, massage certificates, etc. Sounds good? Join us now!

Location: Serbia; Spain Social Discovery Group (SDG) is a group of social discovery companies. SDG solves the problems of loneliness, isolation, and disconnection - transforming virtual intimacy into the new normal. SDG’s products redefine the way people interact and connect with one another. Our portfolio includes social entertainment platforms designed to connect people online across different cultures and regions of the world. We bring together a team of like-minded people and IT professionals who specialize in creating and developing globally impactful social discovery products. Our international team of digital nomads works remotely from all over the world. We’re proud to be a two-time “Great Place to Work” winner (USA & Japan, 2024–2025) and a Top-5 Company for Work-From-Anywhere Jobs (FlexJobs, 2025). We are looking for Lead ML Engineer -LLM Inference & NLP . Your main tasks will be: Speed up and scale LLM inference in production: SGLang, KV and prefix caching, batching, quantization, speculative decoding Run distributed inference for very large models (up to 1T+ parameters) across multi-GPU and multi-node setups Benchmark new GPU servers and hardware, bring them into production and adapt our serving code to them Lead the NLP and CV teams technically: review experiments, set direction, and step in early when something is heading the wrong way Train and fine-tune the language models, and improve the agent harnesses and chat algorithm that run on them Track cutting-edge research and open-source work in inference and post-training, and turn it into the ML roadmap Collaborate closely with the validation, content, and dataset preparation teams to design experiments and measure model quality We expect from you: Deep hands-on experience optimizing LLM inference in production with SGLang, vLLM, or TensorRT-LLM Experience with distributed inference or training of large models: MoE, tensor/expert/pipeline parallelism, multi-node GPU clusters Strong understanding of what makes inference fast: KV cache, attention kernels, batching, quantization, GPU profiling Experience training and fine-tuning LLMs, including post-training (RLHF, DPO, or similar) Proven technical leadership: you've guided engineers through reviews, mentoring and technical decisions while still writing code yourself Proficiency with PyTorch, transformers, and related libraries Experience at AI-focused startups or companies (Character AI, OpenAI, and similar is a strong plus) Backend engineering experience (Python, Go, C#) and knowledge of scalable deployment systems is a significant advantage Advanced English or Russian Nice to have: CUDA or Triton kernel development A computer vision background. We also welcome strong CV leads who have accelerated large generative image or video models Experience with multimodal LLMs First-author papers or notable open-source work, e.g. contributions to SGLang, vLLM, or post-training libraries A degree in CS, math, or physics from a strong program (MSc or PhD) What do we offer: REMOTE OPPORTUNITY to work full-time; The initial pay level or pay range for this role will be shared with candidates during the recruitment process and before the commencement of employment; Vacation 28 calendar days per year; 7 wellness days per year (time off) that can be used to deal with household issues, to lie down and recover without taking sick leave; Bonuses up to $5000 for recommending successful applicants for positions in the company; 50% payment for professional training, international conferences, and meetings; Corporate discount for English lessons; ​ Health benefits. According to the paychecks, if you are not eligible for corporate medical insurance, the company will compensate you with up to $ 1,000 gross per year per employee. This can be spent on self-purchase of health insurance or on doctor’s fees for yourself and close relatives (spouse, children); ​ Workplace organization. The company provides all employees with an equipped workplace and all the necessary equipment (table, armchair, wifi, etc.) in our offices or co-working locations. In the other locations, the company provides reimbursement of workplace costs up to $ 1000 gross once every 3 years, according to the paychecks. This money can be spent on the rent of the co-working room, on equipping the working place at home (desk, chair, Internet, etc.) during those 3 years; Internal gamified gratitude system: receive bonuses from colleagues and exchange them for our merchandise, team building activities, massage certificates, etc. Sounds good? Join us now!

Опыт оптимизации LLM-инференса с SGLang, vLLM или TensorRT-LLM. Работа с большими моделями на распределённой GPU-инфраструктуре.

▫️Разработка и развитие модуля Demand Planning. ▫️Проектирование и реализация высоконагруженных алгоритмов прогнозирования спроса.

Полный цикл разработки и внедрения ML-моделей в продакшен, мониторинг и ретренинг. Лидерство в реализации бизнес-задач.

-поддерживать и улучшать рекомендательный сервис и CV наших продуктов. -участвовать в построении пайплайнов по сбору данных для последующего моделирования.

Задача — обучение с нуля специализированных моделей на 1B–10B параметров. – Подготавливать данные и выстраивать пайплайны обучения.

• Own AI features from concept to production in a high-load environment. • Build and fine-tune LLMs, train specialized models, and integrate them into the platform.

Location: Remote Canada At Affirm, we exist for the moments that matter—giving people a clear, predictable way to pay over time, with no hidden fees, no surprises, and no tradeoffs on what matters most. As a Machine Learning Engineer on our team, you will be at the forefront of developing high-quality, production-ready models that play a central role in our decision-making processes. Your contributions will be instrumental in shaping our financial landscape. If you have a strong interest in machine learning and enjoy challenging work, Affirm is the place for you! What you'll do Use Affirm’s proprietary and other third party data to develop machine learning models that predict the likelihood of default and make an approval or decline decision to achieve business objectives Partner with platform and product engineering teams to build model training, decisioning, and monitoring systems Research ground breaking solutions and develop prototypes that drive the future of credit decisioning at Affirm Implement and scale data pipelines, new features, and algorithms that are essential to our production models Collaborate with the engineering, credit, and product teams to define requirements for new products What we look for 2+ years of experience as a machine learning engineer or PhD in a relevant field Proficiency in machine learning with experience in areas such as Generalized Linear Models, Gradient Boosting, Deep Learning, and Probabilistic Calibration. Domain knowledge in credit risk is a plus Strong engineering skills in Python and data manipulation skills like SQL Experience using large scale distributed systems like Spark or Ray Experience using open source projects and software such as scikit-learn, pandas, NumPy, XGBoost, Kubeflow Experience developing machine learning models at scale from inception to business impact Excellent written and oral communication skills and the capability to drive cross-functional requirements with product and engineering teams The ability to present technical concepts and results in an audience-appropriate way Persistence, patience and a strong sense of responsibility – we build the decision making that enables consumers and partners to place their trust in Affirm! Pay Grade - L Equity Grade - 5 Employees new to Affirm typically come in at the start of the pay range. Affirm focuses on providing a simple and transparent pay structure which is based on a variety of factors, including location, experience and job-related skills. Base pay is part of a total compensation package that may include monthly stipends for health, wellness and tech spending, and benefits (including 100% subsidized medical coverage, dental and vision for you and your dependents). In addition, the employees may be eligible for equity rewards offered by Affirm Holdings, Inc. (parent company). CAN base pay range per year: $133,000 - $183,000 Location - Remote Canada This remote role is open only to candidates residing in Alberta, British Columbia, Manitoba, New Brunswick, Newfoundland and Labrador, Nova Scotia, Ontario, Prince Edward Island, or Saskatchewan. #LI-Remote Remote-first with flexibility built in Affirm is proud to be a remote-first company. Most roles can be done from almost anywhere within the country of employment. Some positions may occasionally require in-person work at an Affirm office, and a few are office-based due to the nature of the work. All new hires will be invited to attend an in-person onboarding experience. Benefits designed for you Our benefits reflect our commitment to care, transparency, and flexibility. Here are a few highlights: Health coverage at no cost: We cover 100% of premiums for employees and their dependents. Spending stipends: Monthly stipends support your tech setup, and the ability to choose health and wellness options that are right for you. Time off to recharge: Flexible time off and generous holiday calendars help you rest when you need to. Own a piece of what you build: Our employee stock purchase plan (ESPP) lets you buy Affirm stock at a discount. We’re committed to providing an inclusive interview process, including accommodations for candidates with disabilities. If you need support, we’re happy to help. For positions based in San Francisco or Los Angeles: Affirm considers qualified applicants with arrest and conviction records, as required by law. By clicking "Submit Application," you acknowledge that you have read Affirm's Global Candidate Privacy Notice and consent to the use of your personal information as described.

Location: Amsterdam, Netherlands; London, United Kingdom; Remote - Europe; Zurich, Switzerland About Nebius: Nebius is leading a new era in cloud infrastructure for the global AI economy. We are building a full-stack AI cloud platform that supports developers and enterprises from data and model training through to production deployment, without the cost and complexity of building large in-house AI/ML infrastructure. Built by engineers, for engineers. From large-scale GPU orchestration to inference optimization, we own the hard problems across compute, storage, networking and applied AI. Listed on Nasdaq (NBIS) and headquartered in Amsterdam, we have a global footprint with R&D hubs across Europe, the UK, North America and Israel. Our team of 1,500+ includes hundreds of engineers with deep expertise across hardware, software and AI R&D. We are seeking a Senior ML Engineer to join a fast-growing team building an agent-native search platform for AI systems, the emerging web access layer for AI. Our focus is agentic search, where AI systems plan, retrieve and refine rather than just return results. You will develop and deploy machine learning models that power retrieval, ranking, and indexing at scale, helping AI systems access fresh, reliable information in real time. This is a high-impact role working on a production system used 24x7, operating at the scale of web search. Your responsibilities: Train and deploy ML models for retrieval, reranking, and search relevance, such as embedding models and cross-encoders, in production Develop models supporting crawling, data selection, and content understanding Build and maintain the data, training and evaluation pipelines behind our models: training-data generation, evaluation sets and retraining Contribute to and improve our ML platform: training and orchestration workflows (Kubernetes, GPU scheduling, experiment tracking, model lifecycle) and inference optimisation for low-latency serving Work closely with engineering to integrate and deploy ML models in production services, balancing latency, quality and cost Apply the latest architectural advances and training procedures for retrieval and LLM-integrated systems Must-haves: 5+ years of experience in software engineering or applied machine learning Strong programming skills in Python and at least one of Go or C++ Proven experience deploying ML models in production systems Solid understanding of machine learning and modern deep learning, including transformers and embeddings Experience building data or ML pipelines and working with workflow orchestration Ability to define meaningful model metrics and evaluate models rigorously Product-oriented mindset with a focus on impact and iteration Nice-to-haves: Hands-on experience with retrieval, ranking, or recommendation Experience with Kubernetes, distributed training or GPU scheduling, or inference optimisation (batching, quantisation, compilation) Contributions to open-source projects, technical publications, or conference talks Strong results on public benchmarks or leaderboards (e.g. MTEB , BEIR ) or models released on Hugging Face Benefits & Perks: Competitive compensation Career growth and learning opportunities Flexibility and ownership Collaborative and innovative culture Opportunity to work on impactful AI projects International environment and talented teams What's it like to work at Nebius: Fast moving - Bold thinking - Constant growth - Meaningful impact - Trust and real ownership - Opportunity to shape the future of AI Equal Opportunity Statement: Nebius is an equal opportunity employer. We are committed to fostering an inclusive and diverse workplace and to providing equal employment opportunities in all aspects of employment. We do not discriminate on the basis of race, color, religion, sex (including pregnancy), national origin, ancestry, age, disability, genetic information, marital status, veteran status, sexual orientation, gender identity or expression, or any other characteristic protected by applicable law. Applicants must be authorized to work in the country in which they apply and will be required to provide proof of employment eligibility as a condition of hire. If you need accommodations during the application process, please let us know.

— Разрабатывать и поддерживать реализованные рекомендательные решения. — Искать новые гипотезы для улучшения пайплайнов обучения моделей и добавления новых признаков.

• Взаимодействие с data scientists, инженерами, аналитиками и бизнес-стейкхолдерами. • Коммерческий опыт в ML/ML Engineering от 3 лет, уровень Middle+ / Senior.

Location: Москва Мы — технологическая команда, создающая инновационные решения для спорта. Наш продукт направлен на трансформацию и объективизацию судейства. Мы ищем опытного специалиста в области MLE для участия в разработке и развитии наших решений. **Предстоит** работать с ключевым функциональным и архитектурным ядром нашей системы видеоаналитики, а именно: Обязанности * работать со скелетом как с многомерным временным рядом: сегментация выступления на элементы и фазы, детекция границ (отрыв, приземление, точки экстремума), выделение повторяющихся циклов, temporal action detection на видеопотоке, сглаживание и агрегация предсказаний по времени * участвовать в проектировании API, внутренних контрактов и структуры данных * оптимизировать модели для достижения целевых показателей задержки * развивать высокопроизводительный Python-фреймворк для CV/ML-пайплайна на более чем 15 моделей, работающих единовременно * реализовывать механизмы обработки с multiprocessing и multithreading. Обеспечивать эффективную оркестрацию между CPU и GPU-ресурсами * минимизировать накладные расходы на межпроцессное взаимодействие, копирование данных и работу с памятью. Проводить профилирование и устранять узкие места в video I/O, decode, inference и postprocessing * проводить профилирование и устранять узкие места на всех этапах: Video I/O, декодирование, препроцессинг, инференс, постобработка. Требования * опыт работы ML Engineer / CV Engineer от 3 лет * высокое качество кода на Python (OOP, Pydantic, SQLAlchemy, для DS - NumPy, OpenCV, PyTorch) * опыт реализации сложной вычислительной или бизнес-логики * практический опыт разработки production-ready ML/CV pipeline * опыт работы с видеоаналитикой, обработкой видео или потоковых данных * понимание особенностей Python runtime для высокопроизводительных пайплайнов: GIL, IPC, serialization, shared memory * опыт использования инструментов ускорения инференса: ONNX, TensorRT, TorchScript или аналогов * опыт работы с Linux, Docker, работы в команде в Git, CI/CD Будет плюсом: * опыт разработки расчетных систем * PostgreSQL, ClickHouse, SQLAlchemy * владение компилируемыми языками, такими как Go или C++, для реализации или оптимизации критичных по производительности частей системы * опыт работы со спортивной аналитикой или системами судейства. Условия * комфортный современный офис рядом с м. Кутузовская  * гибридный формат работы * ежегодный пересмотр зарплаты, квартальная и 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.

Предстоит: Проектировать и внедрять ML-решения (от классического регрессионного анализа до Time Series и Uplift-моделей); Строить высокопроизводительные пайплайны обработки данных и генерации признаков на PySpark;

Location: Austria; France; Germany; Italy; Netherlands; Poland; Spain; United Kingdom Why Kimchi? Great models are raw ingredients. Kimchi is what happens when you ferment them with good engineering. We started inside CAST AI helping companies run LLMs on their own Kubernetes clusters; today we run our own coding harness on any model you pick, and the router deciding which model answers each request is ours too. The Auto Router Our promise is one enforcement layer for your AI: budgets, spend attribution, and model routing, applied before a request ever hits a model. The Auto Router makes the call: for every request, which model serves it. Eight open-source models whose input prices span more than 20×, cached input up to 50× cheaper than fresh, frontier providers on customers' own keys, and a coding agent that fires hundreds of requests before its task is done. Frontier-level task success at a fraction of the token cost: that sentence is the product, and it stays true only if the model inside the router keeps getting better. That model is yours. Made by Cast AI — a Miami-headquartered, globally distributed team across 34 countries that hit unicorn status in January 2026 with a $50B+ backed investment pushing its valuation past $1 billion. About the role Every request gets a decision: which model serves it. This is an applied-ML seat, not an infrastructure one. Platform engineers own the proxy, the streaming path, and the fleet. You own the model inside them: the training data, the objective, the evaluation, and the iteration loop that ships a better router every cycle. The problem: routing is a prediction problem wearing a product's clothes. For every request you predict the quality each candidate model would deliver, what it would cost, and how fast it would decode, and you have to be right enough, in aggregate, that the savings are real and the quality never dips below frontier. The data is messy: rollout outcomes from benchmark harnesses, per-request records from the live fleet, override signals, and token streams whose speed depends on the serving concurrency at the moment. Speed itself is a correctness variable: on deadline-bound agentic work, a model that decodes three times slower doesn't do the same work slower, it does less work. That's why naive cost-per-token objectives quietly optimize for the wrong thing. And the bar for evidence is high: a router change ships only after production-parity evaluation with error bars, then gets held to its numbers in live traffic. Stack: Python; PyTorch; SQL at scale (ClickHouse); OpenAI- and Anthropic-compatible LLM APIs; benchmark and rollout harnesses (RouterBench-class); GitLab CI. Requirements You've spent 5+ years in applied ML or ML research, and you've trained, evaluated, and shipped models that make real-time decisions (ranking, recommendation, bandits, pricing, or LLM routing) and you can tell that story with receipts. Strong Python and SQL at scale: you can pull, clean, and interrogate millions of request-level records without waiting on anyone. You know the difference between a model that got better and an eval that got easier. You have real opinions about what LLM benchmarks measure, agentic harnesses included. You know enough serving reality (TTFT, decode speed, batching, prefix caching, concurrency) to model the objective honestly. You instrument before you optimize, error bars and all. A wide mandate excites you more than it unsettles you. Nice to haves You've worked on LLM routing specifically: RouterBench-class evals, cost/quality/latency Pareto fronts. Calibration, abstention, or conformal methods in production decision systems. You've run online A/B loops with drift monitoring on a model that touches real money. Responsibilities Own the routing model end to end: features, objective, training, calibration, and release cadence. Iterate it like a product, not a research artifact. Build the evaluation that gates every change: production-parity benchmarks, router-vs-router comparisons, holdouts by task family, error bars on pass rates. Nothing ships on vibes. Turn fleet data into training signal: per-request outcomes from production, benchmark rollouts, and override feedback, with failure categorization (infra vs task) so poisoned labels never reach the policy. Model the objective the user actually experiences: task success × token cost × wall-clock speed under the real concurrency regime, with cached-input economics in the picture. Run the online loop: shadow, A/B, full rollout, drift monitoring, retrain. When production disagrees with your offline eval, you find out why before the next ship. Push the frontier where it pays: calibrated decisions, abstention under uncertainty, speed-aware routing. We experiment on what a router should be, not just which model it picks today. Partner with platform engineers on instrumentation: the features and labels recorded per request (decode speed first-class among them) and the experiments the data path must support. What’s in it for you? Competitive salary (depending on the level of experience). Enjoy a flexible, remote-first global environment. Collaborate with a global team of cloud experts and innovators, passionate about pushing the boundaries of Kubernetes technology Equity options. Get quick feedback with a fast-paced workflow. Most feature projects are completed in 1 to 4 weeks. Spend 10% of your work time on personal projects or self-improvement. Learning budget for professional and personal development - including access to international conferences and courses that elevate your skills. Annual hackathon to spark new ideas and strengthen team bonds. Team-building budget and company events to connect with your colleagues. Equipment budget to ensure you have everything you need. Extra days off to help maintain a healthy work-life balance. Hiring process Screening call with Recruiter Hiring Manager interview Technical interview (system design) Live coding Culture Check interview with an executive As part of our standard hiring process, we would like to inform you that a background check may be conducted at the final stage of recruitment through our third-party provider, Checkr. Please note that Cast AI does not provide any form of visa sponsorship/work permit. #LI-Remote

Location: Austria; France; Germany; Italy; Netherlands; Poland; Spain; United Kingdom Why Kimchi? Great models are raw ingredients. Kimchi is what happens when you ferment them with good engineering. We started inside CAST AI helping companies run LLMs on their own Kubernetes clusters; today we run our own coding harness on any model you pick, and we run the models underneath it ourselves. Kimchi Inference One API line and a customer's existing client just works: DeepSeek, GLM, Kimi, MiniMax, Nemotron served serverless from our own GPU fleet (OpenAI- and Anthropic-compatible, pay per token, spend visible per user, team, and key). When control matters, the same models run on the customer's own cluster instead. Behind that one line sits the machinery you'll own: SGLang and vLLM engines, B300 nodes, a shared cluster filesystem that holds every checkpoint exactly once, KV caches riding coding traffic with heavy prefix reuse, and the benchmark harness that gates every change we ship. Made by Cast AI — a Miami-headquartered, globally distributed team across 34 countries that hit unicorn status in January 2026 with a $50B+ backed investment pushing its valuation past $1 billion. About the role Throughput. Latency. Cold starts. Cost per token. Move those four numbers in the right direction and two things happen: customers get faster, cheaper inference, and our margins improve. That's the entire thesis of this role. Every loader you tune, every scheduler decision you land, every kernel you stare at shows up in a customer's p99 and on our P&L. Your mandate is the inference stack's optimization, whatever it takes, and the technical direction of how we optimize it is yours to argue with data. The problem: the checkpoints run 0.3-1.5 TB, and the "right" serving configuration depends on traffic shape, sequence-length distribution, batch dynamics, GPU SKU, memory bandwidth, and quantization tolerance, all variables that shift week to week. Most teams pick a configuration once, over-provision GPUs, and absorb the cost. We don't: we took Kimi-K3 (1.5 TB) from over an hour cold-to-serving to under twelve minutes with a shared filesystem, O_DIRECT weight loaders, and node-local JIT caches, and we're nowhere near done. Prefill/decode disaggregation and tiered KV caches are next. Stack: Python; SGLang; vLLM; TensorRT-LLM; PyTorch; fastsafetensors; CUDA-adjacent tooling; Kubernetes; virtiofs shared filesystems; ClickHouse; OpenTelemetry; Prometheus; Grafana; GitLab CI. Requirements You've built real ML systems for 5+ years, and your portfolio shows depth in inference or training infrastructure, not just model training notebooks. Your Python is production-grade: services, not scripts. You've run vLLM, SGLang, or TensorRT-LLM for real, and you have a working mental model of why an engine performs the way it does on a given GPU. You've measured quantization regressions, not just compression ratios. Multi-GPU and multi-node setups hold no fear: tensor and expert parallelism, collective communication, and their practical failure modes. You know where the bytes live (page caches, O_DIRECT, network filesystems under terabyte-scale reads) and how fast they move. You instrument before you optimize, and you can tell a real win from a benchmark artifact. A wide mandate excites you more than it unsettles you. Nice to haves You've worked on LLM routing specifically: RouterBench-class evals, cost/quality/latency Pareto fronts. Calibration, abstention, or conformal methods in production decision systems. You've run online A/B loops with drift monitoring on a model that touches real money. Responsibilities Push throughput. Continuous batching, chunked prefill, speculative decoding with draft models, CUDA-graph capture, and scheduler tuning across the engine fleet. Find the ceiling on each GPU SKU, then raise it. Cut latency. Attack TTFT and inter-token latency separately; profile to find the actual bottleneck (compute, memory bandwidth, scheduling, storage, or network) and fix it, not the one you assumed. Get more out of the KV cache. Hierarchical and paged attention, prefix caching on heavy-reuse coding traffic, eviction policies, quantized KV. A lot of the unrealized throughput lives here, and you'll own it. Kill cold starts. Weight loaders (O_DIRECT, fastsafetensors-class), JIT and compile caches, shared-filesystem read paths. A 1.5 TB checkpoint has no business taking an hour to reach serving. Quantize without regressing quality. INT8/INT4, FP8/FP4 across weights, activations, and KV, measured on real workloads, not just perplexity benchmarks. Scale across nodes. Distributed inference topologies, network-aware placement, prefill/decode disaggregation, checkpointing strategies that don't bottleneck on storage or interconnect. Set the technical direction. Decide what we benchmark, what we adopt, and what we build ourselves. Bring the team along with strong writeups and reproducible experiments. What’s in it for you? Competitive salary (depending on the level of experience). Enjoy a flexible, remote-first global environment. Collaborate with a global team of cloud experts and innovators, passionate about pushing the boundaries of Kubernetes technology Equity options. Get quick feedback with a fast-paced workflow. Most feature projects are completed in 1 to 4 weeks. Spend 10% of your work time on personal projects or self-improvement. Learning budget for professional and personal development - including access to international conferences and courses that elevate your skills. Annual hackathon to spark new ideas and strengthen team bonds. Team-building budget and company events to connect with your colleagues. Equipment budget to ensure you have everything you need. Extra days off to help maintain a healthy work-life balance. Hiring process Screening call with Recruiter Hiring Manager interview Technical interview (system design) Live coding Culture Check interview with an executive As part of our standard hiring process, we would like to inform you that a background check may be conducted at the final stage of recruitment through our third-party provider, Checkr. Please note that Cast AI does not provide any form of visa sponsorship/work permit. #LI-Remote