Your job is to run careful experiments that show how training data shapes a model's capabilities, its ability to generalize, and its alignment, then turn those findings into guidance the team can use.
Responsibilities
Set up and run controlled supervised fine-tuning (SFT) and reinforcement learning (RL) experiments that measure how datasets change model performance.
Help create public evaluations, benchmarks, and new data types for model research.
Build shared evaluation infrastructure so others can reproduce results across models, capabilities, and domains.
Convert experiment results into advice on data quality, selection, and production.
Write external research, blog posts, and technical reports, and work with partner research teams.
Requirements
0 to 4 years of experience in research or machine learning roles that build or deliver LLM training, evaluation, or post-training systems.
Experience at a frontier AI lab, plus a record of designing, implementing, and analyzing controlled experiments on your own.
Solid understanding of LLM training and evaluation methods, including full post-training experiments.
Skill in designing lightweight experiments and reaching useful conclusions from noisy or incomplete data.
Experience in areas such as finance, software engineering, and policy.
Software engineering skills to build or change evaluation harnesses, training pipelines, benchmarks, or other experiment infrastructure without help.
Research or publications on model evaluation, post-training, or data curation.
Conditions
Base salary ranges from $150,000 to $250,000 per year.
Visa sponsorship is available.
The role is on-site in San Francisco, United States.