Samuel Holt is a Research Scientist at Google DeepMind focused on building LLM agents that learn and act via reinforcement learning, long-context memory, tool use, and inference-time search to enable scalable, assistant-like systems. He holds a PhD in machine learning from Cambridge and an MEng from Oxford, and his first-author publications include NeurIPS, ICLR, and ICML papers with multiple spotlights and a long oral. His work spans memory-augmented agents, unbounded code and content generation, and agents that optimize algorithms and simulators for scientific discovery. Samuel combines rigorous theoretical research with production-minded engineering—having built JAX RL pipelines, sim2real torque-control environments, and high-frequency trading systems earlier in his career. He likes pairing ideas with prototypes and evaluations that push toward self-improving, multi-agent workflows capable of producing large coherent artefacts (e.g., codebases or simulations). Based in London, he’s equally at home discussing planning, long-context generation, RL at inference, or automating parts of the scientific pipeline.
10 years of coding experience
10 years of employment as a software developer
Engineering Science, First Class Honours, Machine Learning, Software, Information, Computer Vision Maters of Engineering (MEng), Engineering Science, First Class Honours, Machine Learning, Software, Information, Computer Vision Maters of Engineering (MEng) at University of Oxford
Doctor of Philosophy - PhD, Machine Learning - LLMs & RL, Doctor of Philosophy - PhD, Machine Learning - LLMs & RL at University of Cambridge
A JavaScript / Python / PHP cryptocurrency trading library with support for more than 85 bitcoin/altcoin exchanges
Contributions:11 commits, 89 PRs, 94 pushes in 13 days
exchangepythonjavascriptphpexchanges
Find and Hire Top DevelopersWe’ve analyzed the programming source code of over 60 million software developers on GitHub and scored them by 50,000 skills. Sign-up on Prog,AI to search for software developers.