Top AI Research Jobs in San Francisco Bay Area, CA

Mercor is offering $5 million in grants for research on AI capabilities, including real-world evaluations, interactive environments, long-horizon tasks, agentic systems, reasoning, post-training, and evaluation methodology. Academic groups, independent researchers, and nonprofits may apply with a focused proposal. Applicants should describe their team, background, research accomplishments, project, expected outputs, impact, timeline, and resource needs. Funded projects are expected to produce public research, datasets, methodologies, or tools.

Lead independent AI research on large language models and agentic systems, from hypothesis through training, evaluation, and production deployment. Develop post-training methods, reasoning and planning capabilities, evaluation benchmarks, synthetic data pipelines, and enterprise AI improvements. Publish at leading conferences, shape model strategy, and mentor researchers and engineers.

Designs benchmarks, reinforcement learning environments, evaluation methodologies, scoring frameworks, and reproducible evaluation architectures for frontier models. Builds and ships production code, maintains data pipelines, analyzes results, and collaborates with research scientists, solutions architects, and ML engineers. The role combines independent research methodology with hands-on engineering across agentic systems, model evaluation, and RL infrastructure.

Leads independent evaluations, adversarial testing, research partnerships, and transparency programs for frontier AI safety. Coordinates researchers, engineers, product, policy, communications, and external experts; translates technical findings into safety recommendations and public documentation; manages goals, milestones, risks, dependencies, and follow-up actions.

Conduct applied AI research focused on computer vision, multimodal, robotics, and coding models. Design datasets, evaluations, benchmarks, and reinforcement learning environments; analyze model failures; develop data remediation specifications; validate data quality through fine-tuning and ablation studies; contribute to open models; publish research; and collaborate with frontier labs to translate model roadmaps into data programs.

Conduct research advancing assurance methods for embodied AI systems, including robots, drones, and autonomous vehicles. Develop prototypes to evaluate safety and reliability methods, define research directions, collaborate with universities, publish scientific findings, and present work at conferences. The role requires machine learning model development, digital tool implementation, and clear communication of complex concepts.

Conduct a 12-week research project using physical-labor video data. Train, evaluate, and ablate models; document findings; collaborate with research, dataset, and evaluation teams; and build required training and data pipelines using PyTorch. The role combines research and engineering and requires in-person work in San Francisco.

Conduct research and develop mathematically rigorous algorithms for real-time auctions, dynamic pricing, bid shaping, pacing, traffic allocation, and fraud detection. Apply machine learning, reinforcement learning, online learning, game theory, forecasting, and Bayesian methods. Deploy models at massive scale, run rapid experiments, evaluate long-term system behavior, collaborate with product and engineering teams, and contribute through publications and internal research seminars.

Join a research and advisory pipeline for an asset-management startup building an auditable reasoning engine. Advise on agentic AI, world models, interpretability, and systemic risk. Help shape research direction, identify failure modes and guardrails, and provide scholarly evidence (papers, GitHub, Google Scholar) and a short written response to demonstrate expertise.

Conduct research on frontier AI agents, including alignment, reliability, long-horizon reasoning, dataset design, evaluation, safety, and scalable infrastructure. Develop novel techniques, run rigorous experiments, document findings in publications and blog posts, and communicate project progress. Work may involve machine learning, deep learning, reinforcement learning, spoken-input agents, and foundation model development.