As a research intern / student researcher, you will collaborate with our research team over a few months on a research project in the domain of coding agents, reinforcement learning or data & evals. The goal of the internship will be to make a meaningful research contribution and publish it as an open-source contribution, a blog or a paper at a top conference.

Here are some example projects you could work on:

• Investigate how we can get richer feedback for reinforcement learning tasks beyond sequence-level credit in the domain of software engineering (e.g. SDPO)

• Come up with a novel method that enables RL task creation at scale (e.g. SWE-Smith) and use it to fine-tune a frontier open source model with our internal training infrastructure

• Build and publish frontier evals measuring coding agents' ability to solve the hardest problems in science & engineering

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 programmers are incredibly well suited to work on the problems we work on.