GenAI Engineer

Location: London Workplace: OnSite About Clarity We’re pioneering Agentic AI — systems that don’t just respond, but reason, act, and adapt autonomously in complex workflows. This is about crafting AI Agent Experiences — designing agents that collaborate seamlessly with humans, learn from context, and make every customer interaction faster, smarter, and more empathetic. You’ll own the technical vision and turn requirements into a live, reliable product used by brands like Grubhub, Booking.com http://Booking.com, Dropbox, Uber, Careem, and Fubo. You’ll collaborate directly with engineers, other tech leads, directors, and the CTO to evolve ambitious prototypes into a rock‑solid, scalable platform What you’ll actually do 50% Build — design & ship - Agentic AI for CX: Real‑time assistants that listen to calls/chats, retrieve from customer KBs, and draft responses with human‑in‑the‑loop controls. - Structured extraction: Schema‑driven pipelines over unstructured text (and other modalities) using retrieval, tool‑use, and robust LLM prompting. - Hybrid anomaly detection: Blend classical time‑series methods (e.g., decomposition, change‑point, forecasting) with LLM‑aware, contextful detectors for seasonality, spikes, step‑changes, and drift. - Novelty discovery: Embedding‑based clustering and drift, topic surfacing, LLM summarization of emerging themes with deduplication and evidence links. - Alerting & scoring: Severity/impact ranking, de‑noising, suppression/cool‑downs, routing, and feedback loops. 25% Architect & scale - Own reliability, latency, and cost. Design online/offline eval harnesses, canaries, and SLAs; operate GPUs/accelerators where needed. - Stand up and harden RAG pipelines (indexing, retrieval policies, grounding, guardrails) and agent frameworks. - Take basic infra ownership on GCP (or AWS/Azure): networking, autoscaling, CI/CD, IaC, observability, and cost tuning. - Participate in on‑call for your area and drive root‑cause analysis with crisp follow‑ups. 15% Collaborate - Pair with back‑end & front‑end to wire extractors/detectors and agents into ticketing, voice, and analytics stacks (APIs, webhooks, real‑time streams). - Partner with PMs/CX to evolve taxonomies, schemas, and guardrails; translate business problems into shipped ML features. 10% Align & showcase - Gather requirements from CX and product leads, demo new capabilities to execs & customers, and document impact with precision/recall, alert quality, latency, and cost metrics. What makes you a great fit - Startup hacker mindset: You self‑start from zero, respect no silos, and carry work from prototype to production. 🛠️ - AI‑native dev tools are your daily drivers: Cursor, v0, Claude Code (or similar). - 7–10 years building production ML/back‑end systems; 2+ years leading while coding. - Expert Python; strong back‑end chops (e.g., FastAPI, gRPC, Postgres, pub/sub/streams). - Agents & RAG: Fluency with at least one agent framework (ADK preferred). Proven track record shipping AI agents and building RAG pipelines. - LLM + DS depth: Prompting/tooling, retrieval design, LLM evals; hands‑on with time‑series analysis (forecasting, change‑point, drift). - Cloud & ops: Basic infra ownership on GCP (or AWS/Azure): networking, autoscaling, CI/CD, IaC, observability, and cost control. - Communication: You explain results clearly, align stakeholders, and write crisp docs. Bonus points - DevOps wizardry; GPU/accelerator experience. - Multimodal pipelines (text + voice + screenshots). - Prior experience in contact center/CX analytics or novelty/anomaly systems. - Founder or founding engineer experience