Research Scientist

Research Scientist - Frontier AI

San Francisco | On-site | $200k–$400k+ + meaningful equity

I’m working with a well-funded frontier AI company in San Francisco building new capabilities in post-training, reinforcement learning, agentic systems and model evaluation.

The company is backed by top-tier investors and has been built by a small team of researchers and engineers from some of the world’s strongest AI and technology organisations. The founding team combines deep research credentials with a track record of shipping large-scale systems, and the business is now entering a major growth phase.

You’ll work directly on how models learn, reason, improve and behave, with the scope to take ideas from first principles through to live model capability.

You may work across areas such as:

• LLM post-training: SFT, RLHF, DPO, GRPO and reinforcement learning
• Agentic systems, tool use and long-horizon reasoning
• Reward modelling and reward design
• RL environments and synthetic data generation
• Benchmarking and novel evaluation methodologies
• Model reasoning, reliability and generalisation
• Training-data strategy and capability improvement
• Research that moves from hypothesis → experiment → training → evaluation → deployment

What they’re looking for

The bar is less about having touched every area above and more about having gone very deep in one or two.

You might have:

• Personally trained or post-trained large language/foundation models
• Built RL environments, reward models or large-scale evaluation systems
• Created benchmarks that exposed meaningful model failure modes
• Worked on agents, reasoning, alignment, safety or continual learning
• Strong Python / PyTorch and hands-on research engineering capability
• Experience running rigorous experiments, ablations and large-scale training
• Published meaningful research, built influential open-source work, or shipped equivalent research inside a leading AI organisation

The strongest people for this team tend to be researchers who can identify an important problem, form a view on how to solve it, run the work themselves and push it all the way through to measurable model improvement.

Why it’s interesting

• You’d be joining while the team is still small enough to have genuine influence over the research agenda.
• Rather than owning one narrow slice of a huge lab, you’ll have the opportunity to work directly with the founders, move quickly across research and engineering, and take real ownership over problems that materially change model capability.
• The company is already well-capitalised, has strong technical backing and is deliberately keeping the team small and high-density.

San Francisco, 5 days on-site.

$200k–$400k+ base depending on level, plus meaningful equity.

Visa sponsorship and relocation can be supported for exceptional candidates.

If your work sits anywhere across post-training, RL, agents, evals, reasoning or foundation models, I’d be keen to speak.

[email protected]