What you'll do
- Design and ship LLM-backed features (RAG, function-calling agents, document understanding) inside studio products
- Build evaluation harnesses so we know whether new prompts and models improved anything before they hit users
- Tune retrieval quality — chunking, embeddings, hybrid search — for the data shape of each product
- Wire in observability and guardrails (PII redaction, prompt injection defense, hallucination scoring)
- Stay current on model releases and ruthlessly evaluate whether they're worth migrating to
What we're looking for
- 3+ years building production systems, with 1+ years shipping LLM-backed features
- Working knowledge of major LLM APIs (Anthropic, OpenAI), embedding models, and at least one vector store
- Strong eval discipline — you don't ship an LLM change without a test set
- Good engineering fundamentals — most of the job is still Python or TypeScript and clean code
- Comfort explaining why a feature should NOT use AI when a deterministic approach is better
What you get
- Equity in what the studio ships — a carry pool across every portfolio company we help raise
- Six weeks paid leave + local public holidays + your birthday off
- A real annual learning budget — no approvals
- Top-decile health, dental, and mental-health cover
- Senior-only team. The people who plan are the people who build.
How we hire
One studio call (45 min), one paid craft conversation, and an offer within 48 hours of the final round — or honest feedback. No coding gauntlets. The full process is on the careers page.