Research Engineer (AI)

TL;DR

Research Engineer (AI/Posttraining): Post-training frontier LLMs to autonomously operate scientific discovery pipelines with an accent on reinforcement learning and agentic environment design. Focus on scaling RL environments, crafting sophisticated reward functions, and collaborating with domain experts to automate experimental design and hypothesis generation.

Location: Based in Menlo Park, California or Remote (USA)

Company

An AI and physical sciences laboratory building frontier models to accelerate novel scientific discoveries.

What you will do

• Post-train frontier models to operate scientific equipment and autonomously run discovery pipelines.
• Scale RL environments and conduct large-scale reinforcement learning experiments.
• Collaborate with physical science experts to define evaluation criteria and build high-quality task benchmarks.
• Design and refine creative reward functions to improve agent performance in complex scientific domains.
• Develop training datasets and evaluation pipelines for AI agents operating in real-world lab settings.

Requirements

• Experience in creating and scaling reinforcement learning environments for LLMs.
• Expertise in defining evaluation criteria, tools, and environments for AI agents in specialized domains.
• Proven track record in training frontier LLMs with RL techniques.
• Skill in crafting training datasets and reward functions using human and AI-based feedback.
• Ability to work closely with cross-functional teams of domain experts.

Culture & Benefits

• Ownership-driven environment with minimal bureaucracy.
• Opportunity to work at the intersection of frontier AI and physical sciences.
• High-growth startup environment with significant autonomy.
• Focus on learning new tools and scientific domains to advance the lab's mission.