As a research engineer, you will work on open-ended research problems in the domain of RL, agents and data, and share your work with the public through open-source releases, papers and talks. We are a small team, which means that you will own problems end-to-end and have to be able to work independently.

These are some of the problems you might tackle on a day-to-day basis:

• Modifying our RL training stack to support end-to-end training with agent harnesses that implement subagents

• Figuring out how we can train agents to inject extremely realistic bugs in codebases, which enables us to build high quality bug-fixing tasks

• Designing experiments allowing us to measure code quality of agent-generated code beyond functional correctness

We care more about your problem solving abilities, creativity, and programming skills than formal research experience or ML fundamentals. In many cases, we have found that creative and high-agency software engineers are incredibly well suited to work on the problems we work on.