The role
Bring rigor to product experimentation. Develop hypotheses, build prototypes, and measure the quality of agent workflows so promising ideas can become reliable capabilities.
What you’ll do
- Investigate planning, code understanding, tool use, and verification for engineering agents.
- Build prototypes and evaluation harnesses with representative tasks.
- Analyze failure modes and distinguish benchmark gains from practical improvements.
- Compare model and workflow choices for quality, latency, and cost.
- Collaborate with product engineers to turn validated research into maintainable implementations.
- Document experiments, limitations, and reproducible results.
What you bring
- Strong programming skills and experience implementing data-driven experiments.
- Familiarity with LLMs, agent workflows, machine learning, or program analysis.
- Ability to design meaningful evaluations and interpret noisy results carefully.
- A solid understanding of software development, testing, and debugging.
- Comfort reading research papers and translating ideas into practical prototypes.
- Clear technical writing and an ability to work independently on uncertain problems.
Nice to have
- Research publications, open-source research tooling, or substantial independent projects.
- Experience with coding-agent evaluation, retrieval, or test generation.
- Applied work on model reliability, security, or inference efficiency.
Your first priorities
- Establish a reproducible baseline for a representative workflow.
- Identify and quantify an important failure mode.
- Deliver an evaluated prototype with a clear recommendation for next steps.
Working remotely
This is a remote role. Eligible hiring locations, working hours, collaboration overlap, and engagement terms will be discussed during the hiring process. We value clear writing, reliable follow-through, and making work visible to teammates.
How to apply
Share your relevant experience, a CV or professional-profile link, and a short example of work that shows how you approach problems. You do not need to match every nice-to-have to introduce yourself.
Read the Privacy Policy before providing personal information. Please do not include sensitive personal information or confidential work from a current or previous employer.