Engineering Manager, ML

San Francisco, United StatesOn-site

Requirements

This is one of the few places at Cursor where infrastructure and model behavior meet directly: when something breaks, it's rarely obvious whether it's a systems bug or the model doing exactly what it was trained to do, and your team has to be good at telling the difference before they can fix it.
YOU MAY BE A FIT IF - You've led engineering teams building infrastructure that trains, evaluates, or serves ML models in production.
- You have strong infrastructure and distributed systems fundamentals: you know what reliability and performance look like under real load, not just in a design doc.
experience with RL training infrastructure, eval frameworks, or building and maintaining simulated environments for model training or testing.

- You can talk fluently with researchers about model behavior and with engineers about systems design, and you know when a problem is actually the other team's. - Bonus: hands-on

The first step in our journey is to build the best tool for professional programmers, using a combination of inventive research, design, and engineering.
Our organization is very flat, and our team is small and talent dense.
We particularly like people who are truth-seeking, passionate, and creative.
We enjoy spirited debate, crazy ideas, and shipping code.
ABOUT THE ROLE You will lead a team of engineers building the infrastructure used to train, test, and evaluate our models.
You'll set technical direction for how we train and evaluate models at scale, stay close enough to the code to debug alongside your team, and work daily with researchers to turn tradeoffs in latency, quality, and cost into infrastructure that actually gets built.
We're hiring across a range of scope for this role, depending on
experience and the size of problem you're ready to own.
- Building the rollout infrastructure that lets researchers run RL experiments at scale without fighting the plumbing.
- Designing eval pipelines that catch regressions before they ship, and give researchers fast, trustworthy signal on whether a change actually helped.