Research Scientist, RL for Autonomous Planning & World Modeling
Looking for a research scientist job in San Francisco? Choose from 68 open roles at 23 employers. Explore real salary data and skills breakdown across remote, hybrid and on-site roles.
Free. No credit card needed.
The roles on this page divide broadly into two clusters. Most are centred on large language models and AI systems: designing evaluations, running pre-training and post-training experiments, improving model behaviour, and building agentic systems. A smaller but distinct group sits in the life sciences, including wet-lab chemistry, computational biology, and AI tooling for scientific research.
For the AI-focused roles, day-to-day work commonly involves running experiments on model capabilities, designing benchmarks, working with reinforcement learning, and contributing to safety or interpretability research. Two of the sample summaries from Databricks — Principal AI Research Scientist, Research Director – AI Scaling and Principal Research Scientist – Scaling — describe leading teams and setting roadmaps for LLM training efficiency.
The life sciences postings at Anthropic range from hands-on chemistry — the Research Scientist, Life Sciences (Chemistry) role involves bench synthesis and characterisation — to computational pipeline work in the Research Scientist, Life Sciences (Computational) posting, to building agentic tools for scientific tasks in the Research Scientist, Life Sciences role.
Skills counts reflect how often a term appears in posting descriptions across this sample of postings, not whether each skill is strictly required. Evaluation is the most frequently mentioned term, followed by LLMs and reinforcement learning. PhD and publications appear in roughly half the postings, signalling that academic research backgrounds are commonly referenced, though the framing varies by role and employer. Python and PyTorch appear at similar rates.
The employers' own advertised ranges, for individual roles at different levels — not an average and not a market rate.
The majority of postings on this page are listed as onsite, meaning a physical San Francisco presence is expected. A notable share are hybrid, though the specific in-office cadence is not stated in most cases. Only a very small number are listed as fully remote. No postings have an unknown work mode. Candidates should treat each posting individually, as hybrid terms differ by employer and are not standardised across this set.
Jobs checked 2 hours ago. · Guide reviewed 9 September 2026.
Search by role, company, or anything a posting mentions.
