Staff Research Engineer

About the role

Cursor is hiring a Staff Research Engineer to work at the core of how the company turns AI capability into a product people rely on. This is not a research-only or a glue-code role — you'll own research engineer systems end to end, from prototype through evaluation, rollout, and the iteration that follows once real users are in the loop.

You'll sit close to the product surface, partnering with engineering, product, and research to ship features that depend on model behaviour being reliable, fast, and cost-effective. The hard part of this job is rarely the first demo; it's making something work consistently in production, measuring whether it actually helps, and improving it. That's the work we're hiring for.

The role is based in San Francisco, CA. It's a staff-level position on a team that ships continuously.

What you'll do

• Design, build, and ship research engineer systems in production — owning them end to end rather than handing prototypes off.
• Take PyTorch and RL work from idea to launch, including the evaluation harness that proves it works.
• Partner with product and research to turn model capability into reliable, measurable features.
• Own the quality, latency, and cost of what you ship, and keep improving all three after launch.
• Write clearly — design docs, postmortems, and the reasoning behind the calls you make under model-driven ambiguity.
• Set technical direction for your area and raise the bar for engineers around you.

What we're looking for

• Demonstrated experience shipping PyTorch and RL in production — not just notebooks or side projects.
• Strong engineering fundamentals and the judgment to make decisions when the model behaviour, not the spec, is the hard part.
• Comfort across the stack of an AI product, from data and models to the surface users touch.
• Working familiarity with Pretraining and Evals.
• A track record of owning systems end to end and mentoring other engineers.
• Clear written and verbal communication — Cursor moves fast and writes a lot.

Bonus points

• You've shipped something AI-native that real users depend on, and can talk through what broke and how you fixed it.
• Open-source contributions, a strong portfolio, or published work in or around research engineer.
• Experience at an earlier-stage company where you wore more than one hat.

About Cursor

Cursor is an AI-native company — AI isn't a feature bolted onto the side, it's the core of what the product does. That changes the work: the systems you build compound directly into the company's value, which is exactly why Research Engineers here are paid and trusted accordingly.

The team is small enough that your work is visible and large enough that you'll have specialists to learn from. If you want to do the highest-leverage research engineer work of your career alongside people who hold a high bar, this is the kind of place built for it. See more open roles and culture signals on the Cursor hub.

Compensation & benefits

The posted total-compensation band for this role is $290k–$380k, combining base salary and equity. Equity is a meaningful slice at this level, so the realised number tracks how the company performs over your vest. For how this compares across companies, see the Research Engineer salary guide.

• Meaningful equity with a standard four-year vest.
• Comprehensive health, dental, and vision coverage.
• On-site or hybrid, with a flexible schedule.
• Annual learning-and-development budget and top-tier hardware.
• Generous paid time off and parental leave.

How the interview works

AI-native loops look a little different from traditional ones. A typical process for this role runs:

• Recruiter screen — background, motivation, and a comp/level calibration.
• Technical screen — a practical coding or modelling exercise grounded in real work.
• Deep dive — a system or applied-AI session on how you'd design and evaluate research engineer in production.
• Team & values — collaboration, communication, and how you handle ambiguity.
• Final — a wrap-up with a hiring manager or founder.

Your level — and therefore your offer — is set in this loop, which is why preparation matters more than negotiation afterwards. Landed maps your background to this exact posting, shows you the skills and signals you're missing, and drills the interview until you walk in ready.