Job Summary
WHAT YOU WILL OWN
Features across the agent runtime, RAG, memory, evaluation and the Studio, working inside a senior engineer's track. You will build the reference agents, the knowledge-base pipeline, the eval datasets and much of the pilot feedback work — the parts of the product where AI quality is decided.
MUST HAVE
-
Strong Python with modern tooling (typing, Pydantic, async, pytest); can work in TypeScript / React when needed.
-
Hands-on with LLM APIs: tool / function calling, structured outputs, prompt iteration, token and cost awareness; has shipped at least one LLM-powered feature to real users.
-
Built a RAG pipeline or search feature: chunking, embeddings, a vector database, retrieval evaluation.
-
Comfortable with PostgreSQL and Redis; understands REST API design and event-driven basics.
-
Works with Docker, Git and CI daily; writes tests without being asked.
-
Curious and rigorous: reads model behaviour from traces, forms a hypothesis, tests it.
GOOD TO HAVE
-
LangGraph or another agent framework. Temporal. Kubernetes. Experience building evaluation datasets or with LLM-as-judge. Prior work on enterprise integrations.
Key Responsibilities
RESPONSIBILITIES
-
Implement features end to end — API, data model, tests, telemetry — from a design agreed with your track lead.
-
Build and tune agents: prompts, tool schemas, structured outputs, routing rules; measure them with the eval harness rather than by feel.
-
Own the knowledge-base pipeline (ingestion, chunking, embedding, retrieval quality) and the memory features.
-
Turn pilot transcripts and traces into eval cases and fixes.
-
Contribute to the Studio (React / TypeScript) when your feature needs a UI.
Skill Requirements
MUST HAVE
-
Strong Python with modern tooling (typing, Pydantic, async, pytest); can work in TypeScript / React when needed.
-
Hands-on with LLM APIs: tool / function calling, structured outputs, prompt iteration, token and cost awareness; has shipped at least one LLM-powered feature to real users.
-
Built a RAG pipeline or search feature: chunking, embeddings, a vector database, retrieval evaluation.
-
Comfortable with PostgreSQL and Redis; understands REST API design and event-driven basics.
-
Works with Docker, Git and CI daily; writes tests without being asked.
-
Curious and rigorous: reads model behaviour from traces, forms a hypothesis, tests it.