Take AI features from prototype to production, with evaluation, guardrails and costs that hold up.
- Type
- Full-time
- Location
- Remote
- Level
- Mid–Senior
- Team
- Engineering
About the role
You will build retrieval systems, agent workflows and evaluation harnesses that go into real products. The interesting work here is not the demo; it is everything required to trust the demo in production.
What you'll do
- Design and tune retrieval pipelines over client corpora
- Build agent workflows with typed tools and hard boundaries
- Create golden datasets and automated evaluation in CI
- Implement guardrails, logging and cost controls
- Work with clients on data governance and privacy constraints
What we're looking for
- 3+ years in software engineering, with recent production LLM work
- Strong Python or TypeScript
- Practical experience with embeddings, vector search and reranking
- Rigour about evaluation: opinions are not metrics
- Understanding of prompt injection and data-handling risk
Nice to have
- Experience self-hosting open-weight models
- Background in information retrieval
- Familiarity with data pipeline tooling
How to apply
Send work, not a cover letter.
Email us something you built and two or three sentences on what you decided and why. A repository, a live URL or a case study all work. No portfolio PDF required.
- Response time
- Within 5 working days
- Work sample
- Paid at your rate
- Decision
- Within a week of the last call
Two to three weeks end to end, and a decision either way within a week of the last conversation.
- 01
Intro conversation
Thirty minutes on what you have built and what you want next. No trick questions, no whiteboard.
- 02
Craft conversation
A deep dive into work you are proud of: decisions, trade-offs and what you would do differently.
- 03
Paid work sample
A small, realistic piece of work, scoped to a few hours and paid at your rate. We never ask for free labour.
- 04
Team & offer
Meet the people you would work with, ask us anything, and get a decision within a week.
Curious about the team, the projects or how the work sample runs? Email us; a real person answers, usually the one you'd be working with.