Compound AI Systems | Agents | Retrieval | Evaluation Infrastructure | Deployment | Voice
London (hybrid, Tue–Thu) + up to 30% UK/EU client travel
£150–200k + equity & benefits
This is a production LLM systems, evaluation infrastructure, agent architecture, enterprise deployment role. Backend engineers who moved into AI at a lab, an applied-AI company or a Big Tech applied team and have shipped an LLM or agent system to real users, measured it in production and changed it when the numbers said so. You will be tasked with building one production AI system at a time for a Fortune 500 client inside a $1.8B applied-AI company (ex-Palantir founders, $202M raised) opening its London office: scope it with the client’s senior stakeholders, build it on the company’s internal LLM application platform, prove it with evals, run it, and feed what you learn back into the platform.
Own the system: one production AI system live at the London client in six months, with evals, observability and a client lead who trusts it.
Own the measure: an eval set built from real traffic, a published accuracy and cost baseline, and every change gated on it.
Own the room: scope the next two builds directly with the client’s senior stakeholders, and ship reusable components back into the platform.
THE ROLE:
- Production AI systems: agents, tool use, retrieval, prompting and the full-stack product around them; from prototype to a system a Fortune 500 runs
- Evaluation infrastructure: eval sets built from production traffic, segmentation by case type, LLM-as-judge with independent scoring, confidence intervals, regression gates before any prompt, model or routing change ships
- Autonomous loops: the shift from deterministic workflows to agentic systems that self-improve from transcripts, tickets and NPS; versioning of every component so configurations can be compared
- Enterprise estates: integration with client CRMs, EMRs, telco back ends and data platforms; security reviews, SOC 2, data residency; cost and latency ceilings
- Platform contribution: reusable infrastructure, tools and workflows returned to the internal LLM application platform for use across deployments
- Voice and real-time: streaming pipelines, speech models, model fallback and routing under rate limits
- Client ownership: 60% building, 40% with the client; scoping with C-level stakeholders, pushing back when the brief is wrong, reframing the problem
- AI-native workflow: coding agents (Claude Code) used daily, with judgement about what they produce
- Languages and tools: Python first; TypeScript for product surfaces; AWS, GCP or Azure; evals, observability and CI/CD as standard
- Delivery: one 3-month build at a time, two or three concurrent across a long client relationship; works with AI strategists, architects, a Founding Data Scientist and the Head of Engineering
Seniority: Senior / Staff (IC)
Industry: Applied AI, Enterprise Software, AI Transformation
Employment Type: Full-time, permanent
Skills: Python, LLM systems, AI agents, Tool use, RAG and retrieval, Evaluation frameworks, LLM-as-judge, Observability, Prompt engineering, System architecture, Enterprise integration, Streaming and voice, Claude Code, AWS/GCP/Azure, Client-facing delivery
#J-18808-Ljbffr
Artificial Intelligence Engineer employer: Intelix.AI
Join a leading global AI and analytics firm that champions innovation at the intersection of knowledge graphs and generative AI. With a strong emphasis on employee growth, you will have the opportunity to work in a dynamic, hybrid environment that fosters collaboration and creativity while tackling high-impact projects for prestigious clients. Enjoy competitive compensation, meaningful equity options, and a culture that values autonomy and continuous learning, making it an exceptional place for data professionals to thrive.
