Location: King’s Cross, London (in-office) | Experience: 5+ years | Openings: 4
About my client
My client is building an AI work-execution platform for alternative investment firms and their portfolio companies, securely bringing together proprietary data, external sources and workflows to deliver faster, more contextual analysis and better decisions. Backed by a global group of 1,800+ professionals with deep expertise in investment diligence, research, valuation and fund administration, the UK core team is small, senior and high-ownership, with direct access to the Executive Committee.
The role
You will design, build and evaluate the LLM-powered systems at the heart of the platform, working in Python/FastAPI with LangGraph and LangChain across leading model providers. Token cost, latency and auditability are first-class engineering constraints. Senior engineers own major capabilities. Lead engineers also set technical direction and mentor.
Key responsibilities
- Build multi-step agentic workflows (LangGraph / LangChain) with tool-calling, checkpoints and human-in-the-loop patterns
- Integrate Anthropic, OpenAI and Amazon Bedrock models, owning routing, fallback and context and cost management
- Build structured-output systems and MCP tool servers for reasoning over investment documents
- Own production RAG pipelines (Qdrant, reranking, hybrid search) and extend GraphRAG on Neo4j
- Run evaluations (golden datasets, LLM-as-judge, regression gates) and instrument every model call for latency, cost and quality
- Apply responsible-AI guardrails in a UK/GDPR context, shipping on FastAPI, PostgreSQL and AWS with CI/CD
- 5+ years of software engineering with strong Python, including 2+ years running LLM or ML systems in production
- Deep hands-on experience with an agent framework (LangGraph, LangChain or similar) beyond prototypes
- Production RAG experience: embeddings, vector databases, chunking and retrieval evaluation
- Graph-retrieval experience, or the data-modelling fundamentals to pick up Neo4j and Cypher quickly
- Evaluation-led development: you can describe the eval harness of your last system from memory
- Daily use of AI coding tools (Cursor, Claude Code or similar), with a considered view of their strengths and limits
Nice to have:
Qdrant, Weaviate or pgvector in production; MCP servers; LangSmith / LangFuse / Ragas; fine-tuning or open-weight deployment; financial-services documents.
Education:
Bachelor’s degree in engineering, computer science or a related field, or equivalent experience.
What’s on offer
- Close collaboration with an Executive Committee with 20+ years’ experience serving 400+ leading alternative investment clients
- Competitive compensation
- The chance to build something impactful for leading financial services firms
- A supportive, collaborative environment in a central London office
#J-18808-Ljbffr…