Ema
AI Resident
San Francisco Bay Area · Full-time · Posted Jul 31, 2026
- Education
- No specific degree required; strong undergraduates, graduate students, and self-taught builders are welcome.
About the job
Own an AI research problem end to end, from proposal and implementation through evaluation, gated production launch, and a write-up of findings.
What you'll do
- Build and ship a system for a scoped problem in agent training, evaluation, data, or efficiency.
- Design experiments and evaluations using production traces and data.
- Document results, including what did not work.
What they ask for
- Demonstrated depth in ML or agent systems.
- Solid ML fundamentals and strong engineering; Python and PyTorch.
- Depth in at least one area: post-training, reward modeling or LLM judges, agent and tool-use systems, retrieval and memory, or evaluation design.
- Statistical literacy and a habit of honest measurement.
Nice to have
- Hands-on post-training with open models, or experience with interactive agent environments.
- Experience in trace analysis, serving and efficiency, multi-node GPU training, publications, open-source work, or security.
Pay
$4,000 month (salary; currency not stated)
Tools
Python, PyTorch, TRL, veRL, OpenRLHF, vLLM, SGLang
AI focus
Research and build production agent systems, including post-training, agent harnesses, data and evaluation pipelines, and efficiency improvements.
Model-written summary · the original posting is authoritative