Job description
ABOUT THE COMPANY We're building autonomous research agents for recursive self-improvement (multi-agent systems that propose, run, and analyze machine learning experiments). We're a small team based in San Francisco, on-site ABOUT THE ROLE You'll build and maintain the research systems and pipelines that our research runs on top of: data pipelines, training infrastructure, evaluation tooling, deployment, observability.
The work bridges research and production, and you'll be the person who makes "we ran an experiment" actually mean "we ran it correctly, at scale, with results we trust." You'll own systems end-to-end. You'll work with researchers daily and translate research code into infrastructure that the team can rely on. You'll move fast and you'll be measured on whether your systems make the team faster.
WHAT YOU'LL DO
- Build and maintain the training, evaluation, and deployment pipelines that our research runs on
- Take research code from prototype to production: refactor, harden, instrument, test
- Design observability into our research systems (metrics, logs, traces, eval dashboards) so failures surface fast
- Own data pipelines for training and evaluation: ingest, dedup, version, validate
- Work closely with researchers to understand what they need, what's slow, and what's brittle
- Set engineering standards across our research stack (testing, reviews, runbooks) so the team scales
- Contribute to architectural decisions that shape how research and production interact WHAT WE'RE LOOKING FOR
- Senior research engineer with 6+ years building production-grade research systems
- Track record across the full lifecycle: data, training, evaluation, deployment, monitoring
- Strong distributed systems experience; you've shipped systems that have to be on
- Fluent Python, fluent with at least one of (Py. Torch, JAX); comfortable at the systems-level when needed
- Comfortable with experimentation infrastructure (Ray, Slurm, Kubernetes, or similar)
- Bias toward shipping; you prefer working code over working diagrams
- Strong written communication NICE TO HAVE
- Experience building experimentation platforms or research infrastructure at a frontier research lab
- Background in distributed training systems
- Open-source contributions to research infrastructure
- History of working effectively with small senior teams THIS ROLE IS PROBABLY NOT FOR YOU IF
- You want to do research with engineering as a side activity: this is engineering as the main thing
- Cross-functional work with researchers (translation, scoping, education) doesn't appeal
- Long-running ownership of running systems isn't appealing: this role has it