Software Engineer

We are seeking a software engineer to help design and build challenging evaluation environments for advanced AI systems. The work focuses on creating realistic technical tasks that test how well agents handle complex engineering problems, including multi-step workflows, large codebases, ambiguous requirements, and system design decisions. This is a hands-on role that combines product-minded engineering with research-driven iteration. You will develop tasks from the first idea through implementation, scoring, debugging, and refinement. A big part of the job is understanding where an automated agent is genuinely limited versus where the environment or evaluation logic needs improvement, then adjusting the setup to better measure the desired behavior. You will work independently on high-impact problems, collaborate closely with a small engineering team, and contribute to internal tools and infrastructure that support the broader environment-building effort. The role also involves mentoring as the team expands. We are looking for someone with broad software engineering experience and deep strength in at least one area such as infrastructure, distributed systems, performance, security, compilers, or databases. Strong Python skills are important, along with substantial experience using coding agents or similar developer tools. You should be comfortable spotting where these tools take shortcuts and directing them effectively. Prior experience owning complex work end-to-end and thriving in ambiguous settings is essential.