Research Scientists & Research Engineers, Foundation models

Research Scientists \& Research Engineers – Foundation Models

Ready to help build what comes after the transformer?

Most frontier language models are still constrained by architectures that become increasingly expensive as context grows. Longer reasoning, persistent state, and truly agentic behaviour remain difficult problems.

This team is taking a different approach.

They're building a new generation of foundation models designed for extremely long-context reasoning and persistent state, rethinking how models scale across both training and inference. Rather than iterating on today's architectures, they're developing fundamentally new approaches that make much larger context windows practical while improving efficiency.

The research team is expanding across several areas as they scale their models to create different sizes

20B to XXXB\+ parameters

, training on

hundreds of billions through to trillions of tokens

.

Depending on your background, you'll work across one of three areas:

Research Scientist – Pre-Training

Design and train large-scale foundation models from the ground up, working across pre-training strategy, large-scale datasets, optimisation, training dynamics, and long-context behaviour. You'll help answer fundamental questions around how these new architectures learn and scale.

Research Scientist – Post-Training \& Reinforcement Learning

Develop the methods that improve model capability after pre-training. The work spans reinforcement learning, reasoning, evaluation, long-horizon behaviour, and large-scale experimentation, helping models plan, use tools, and operate reliably across complex environments.

Research Engineer – Large-Scale Training Systems

Build the systems that make frontier research possible. You'll work across distributed training, GPU optimisation, training infrastructure, experimentation platforms, inference, evaluation pipelines, and ML systems that enable researchers to iterate at scale.

Across every team you'll work with a small group of highly technical researchers and engineers, with significant compute resources and the freedom to tackle difficult research problems from first principles.

Whether your expertise is large-scale pre-training, reinforcement learning, distributed systems, or research infrastructure, you'll have the opportunity to influence both the research direction and the systems being built.

Package

• Remote across the US, with hybrid opportunities in San Francisco
• Base salary up to approximately

$400,000

• Significant equity package
• Compensation negotiable depending on experience

All applicants will receive a response.