Artificial Intelligence Researcher

AI Researcher - Reinforcement Learning & Post-Training

$300,000–$500,000 base + meaningful equity

New York, San Francisco or Seattle | Hybrid, three days in office

We’re partnering with a well-funded, early-stage AI research company working directly with leading frontier labs on some of their hardest reinforcement learning and post-training problems.

The team develops the environments, trajectory datasets and training signals that help models acquire new capabilities. The research centres on a practical question: what makes a model meaningfully better, and how do you prove it?

Joining a small applied research team, you’ll work directly with customer researchers to turn open-ended problems into experiments, training methods and measurable improvements. You’ll have scope to shape the research approach and influence what the team investigates next.

The work

• Design and run post-training experiments using reinforcement learning, supervised fine-tuning and preference optimization.
• Develop reward functions, verifiers and evaluation frameworks that distinguish real capability gains from reward hacking.
• Shape training environments and trajectory data, investigating how correctness, coverage and data quality affect model behavior.
• Analyze training dynamics and experimental results across distributed GPU workloads, using the findings to guide subsequent research.
• Collaborate directly with frontier research teams, translating ambitious research objectives into concrete technical deliverables.

What we’re looking for

You have hands-on experience in AI research and can explain the models you trained, the decisions you made and the results you achieved.

Your background will likely include:

• 2-7+ years in applied research or a research scientist role, with substantive reinforcement learning experience.
• Model-training work at a frontier lab, a respected emerging research lab or an AI data lab—or published research or a released model with strong results on a recognized benchmark.
• Strong experimental judgement: knowing which signals to trust, where evaluations fall short and what to investigate next.
• Practical experience with Python, PyTorch and modern model-training workflows.

An RL-focused PhD is ideal, but demonstrated research ability and results matter more than the qualification alone.

Why consider it?

This is an opportunity to combine direct exposure to frontier research problems with the ownership of a small, growing team. Research depth is central to the role, and your work will directly inform the environments and data used to train advanced AI systems.

The package includes $300,000–$500,000 base salary and equity, with visa sponsorship and transfers supported.

To explore the opportunity, share your CV or research profile, ideally with a paper, model release or brief example of training work you personally owned.