Sr. Model Engineer

About this role

Drive research and development of frontier coding and computer-use agents, with a focus on reinforcement learning and mid-training. Own research problems end to end, from hypotheses and experiments through evaluation and scaling promising results into training runs.

About the company

Cursor builds an AI coding editor and agent workflows that help developers write and ship software faster. It serves software teams looking for tighter AI assistance inside the IDE.

The team

Machine Learning

Stack

What you'll do

• Design rigorous experiments to improve model capabilities.
• Improve training data quality, deduplication, and mixture recipes.
• Develop RL data, rewards, and graders while preventing reward hacking.
• Advance model training and post-training recipes for coding and computer-use agents.
• Build evaluations and targeted tasks to identify model weaknesses.
• Scale promising research results into frontier-scale training runs.

What we're looking for

• 3+ years of hands-on experience training or post-training neural networks with 1B+ parameters.
• Experience with pre-training, fine-tuning, or RL methods such as RLHF or DPO.
• Production-quality programming experience in Python, PyTorch, JAX, Rust, or C++.
• Able to work onsite in Palo Alto.

Nice to have

• Experience with language-model RL algorithms, scaling laws, or RL environments and evaluations.
• Distributed training experience and skill debugging issues at scale.
• Experience training frontier-scale models.
• Strong RL and machine learning fundamentals.
• Original ML contributions demonstrated through research, open source, or production impact.
• MS or PhD in a related field, or equivalent research experience.

Benefits

• Equity
• Medical, vision, and dental coverage
• 401(k) retirement plan
• Short- and long-term disability insurance
• Life insurance