Product Engineer
Build and ship products in the AI space, from the data model and the runtime to the screen a customer touches.
You will own features end to end. The schema, the service, the failure modes, and the interface. Not a ticket queue, and not a slice of someone else's design. Much of the work sits behind a conversational product: client state, realtime protocols, durable scheduling, and a runtime that stays dependable when networks and model providers fail. Calling a language model is the easy part. Making it reliable is the job.
- Design and build features across the stack, from schema to interface.
- Make systems survive real conditions. Reconnection, retries, partial failure, recovery.
- Choose the data structures and interfaces that keep the next change cheap.
- Write focused tests for the behaviour that matters, not for a coverage number.
- Ship to real users on a short loop and act on what comes back.
- Explain your decisions, your trade-offs, and the scope you left out.
We care less about years of experience than about evidence. What you have shipped, the complexity you handled, and how you decide.
- You have shipped something real and can explain every part of it.
- Strong fundamentals in data modelling, concurrency, and failure handling.
- Comfortable across the stack, even if you are stronger on one side.
- You reach for the simplest thing that actually works.
- You write clearly. Most engineering disagreements get resolved in writing.
- You handle ambiguity by making a reasonable assumption, documenting it, and continuing.
- 01The challenge
Fork the public repo, choose one of five focused problems, and spend roughly 6 to 8 hours on it.
- 02Technical conversation
We walk through your submission together. Architecture, the failure cases you handled, and one trade-off you made.
- 03Team conversation
How you work, what you want to build next, and whatever you want to ask us. Both sides decide.
Every submission is scored on the same public scorecard, whatever stack you choose. Core correctness, architecture, coding patterns, failure handling, testing, and communication. You can read it before you start.
- TypeScript
- React
- Next.js
- React Native
- Postgres
- Realtime
- Durable workflows
- Testing
Apply by building
Fork the repo, pick one of the five problems, and send us the link. About 6 to 8 hours of work, due within 3 to 4 days.
Not the right role, but the right team?