Research Scientist

Prime Recruitment Partners; San Francisco (CA); On-site; USD 140,000 - 230,000; Full time

Job summary

Prime Recruitment Partners is partnering with a well-funded, early-stage AI lab to advance fundamental research at the intersection of reinforcement learning, reasoning, and computational problems.

You will build and train RL agents, design curricula and reward structures, and develop evaluation frameworks to separate genuine reasoning from mere benchmark optimization. The role scales training across distributed GPU infrastructure and collaborates with world-class researchers.

Qualifications

• PhD in Machine Learning, Computer Science, or related quantitative field.
• Hands-on experience building reinforcement learning systems and training large-scale models.
• Strong publication record at top-tier conferences (NeurIPS, ICML, ICLR, or equivalent).
• Expertise with PyTorch or JAX and distributed training.

Responsibilities

• Build and train reinforcement learning agents capable of solving complex technical reasoning problems.
• Design reward functions, training curricula, and action spaces for long-horizon learning and discovery.
• Develop evaluation frameworks that distinguish genuine reasoning from benchmark optimization.
• Scale training across distributed GPU infrastructure using modern ML frameworks and tooling.
• Work alongside domain experts to take research ideas from experimentation through production.
• Publish impactful research at leading ML conferences.

Skills

Reinforcement learning

Distributed training

Research publication

Collaborative research

Education

PhD in ML/CS

Tools

PyTorch

JAX

Job description

We're partnering with a well-funded, early-stage AI lab building research systems aimed at one of the hardest open problems in the field: getting models to do real discovery work rather than pattern‑match their way through benchmarks. The team is drawn from leading AI labs and top research institutions. This is not product engineering or benchmark chasing. It's fundamental research at the intersection of reinforcement learning, reasoning, and computational research problems, in an environment built around exploration, publication, and close collaboration with world‑class researchers and engineers.

What you'll be working on:

• Build and train reinforcement learning agents capable of solving complex technical reasoning problems.
• Design reward functions, training curricula, and action spaces for long‑horizon learning and discovery.
• Develop evaluation frameworks that distinguish genuine reasoning from benchmark optimization.
• Scale training across distributed GPU infrastructure using modern ML frameworks and tooling.
• Work alongside domain experts to take research ideas from experimentation through production.
• Publish impactful research at leading ML conferences.
• PhD in Machine Learning, Computer Science, or a related quantitative field.
• Strong publication record at top‑tier conferences (NeurIPS, ICML, ICLR, or equivalent).
• Hands‑on experience building reinforcement learning systems and training large‑scale models.
• Expertise with PyTorch or JAX and distributed training.
• Experience applying RL to mathematics, coding, simulation, or other complex reasoning problems is a strong plus.