Research Engineer, Synthetic Data (AI)

Location: San Francisco, United States; onsite work and an onsite interview at the office are required.

Company

Cursor AI builds tools and research systems aimed at automating coding for professional programmers.

What you will do

• Design end-to-end synthetic data recipes using traces, documents, codebases, and workflows.
• Build scalable data generation pipelines, verification systems, and anti-hacking mechanisms.
• Control task difficulty and dataset diversity so generated data reflects real usage.
• Run ablations and analyses to demonstrate that datasets improve model capabilities.
• Use traces and training results to address capability gaps and turn successful recipes into reusable infrastructure.
• Partner with research and reinforcement learning data teams to translate model goals into synthetic data plans.

Requirements

• Research experience with synthetic or post-training data in industry or academia.
• Experience designing data generation methods, deploying data into training, and demonstrating model impact.
• Strong software engineering fundamentals and experience building substantial data pipelines, agent systems, or evaluation frameworks.
• Ability to design training tasks that are useful, reliably gradable, and resistant to reward hacking.
• Careful, systematic analysis of aggregate metrics and raw model outputs.
• Ability to work onsite in San Francisco and attend an onsite interview at the office.

Culture & Benefits

• Small, flat, talent-dense organization.
• Work environment encourages truth-seeking, creativity, spirited debate, and unconventional ideas.
• Close collaboration with research and reinforcement learning data teams.

Hiring process

• Two or three short technical interviews.
• Onsite interview including a small project, discussion of ideas, and team meetings.