Research Engineer Synthetic Data
Anysphere, Inc. (doing business as Cursor)Visit Anysphere, Inc. (doing business as Cursor) website
Cursor is an AI coding-agent product operated by Anysphere, Inc.; it was acquired by SpaceX in August 2026 and continues to operate.
Distributed
Funding history
About Anysphere, Inc. (doing business as Cursor)
Cursor provides AI-assisted software-development tools, including desktop and CLI coding agents, autonomous cloud agents, code review, and enterprise offerings.
Skills
About the Role
You will identify missing model capabilities, develop synthetic-data generation strategies, and build scalable pipelines and verifiers. You will run ablations, inspect training signals and traces, validate capability gains, improve successful recipes, and translate capability goals into concrete data plans.
Requirements
- Research experience with synthetic or post-training data
- Experience designing a data-generation method, deploying data to training, and demonstrating model impact
- Strong software engineering fundamentals
- Experience building data pipelines, agent systems, or evaluation frameworks
- Knowledge of training-task design and reward-hacking resistance
- Ability to analyze aggregate metrics and raw outputs
Responsibilities
- Design end-to-end synthetic-data recipes from raw sources
- Build verification and anti-hacking mechanisms
- Control task difficulty and diversity
- Run ablations and analyses to validate capability improvements
- Iterate from traces and training results to address capability gaps
- Convert successful recipes into reusable infrastructure
- Translate model capability goals into synthetic-data plans
Hiring Process
2–3 short technical interviews, followed by an onsite office visit involving a small project, idea discussion, and team meetings.