I am an AI researcher working on frontier models, AI security & safety, robustness, interpretability, and understanding intelligence. I am Vice President, AI Security at Cohere, where I lead global AI security research and work with the core model team on frontier-model development and training.
I co-founded AISLE, an AI security research startup acquired by Cohere in November 2026. As its CTO and Chief Scientist, I led AISLE's AI research and engineering. Our team built autonomous AI systems that discovered more than 500 previously unknown vulnerabilities across over 100 codebases, including all 12 CVEs in OpenSSL's January 2026 security release, and often outperformed frontier AI models in detecting zero-day vulnerabilities in some of the most secure codebases.
I am also a theoretical physicist by training, a fan of progress, and an optimist about technological solutions to our biggest problems. You can find my academic research on Google Scholar.
Selected work
- AISLE: 500 CVEs in one year
- Scaling laws for language-model activation attacks
- Deep Ensembles: A Loss Landscape Perspective
- Training a Helpful and Harmless Assistant with RLHF
Some professional things about me
- ๐ Vice President, AI Security at Cohere, leading AI security research and working on frontier models
- ๐ Co-founded AISLE and served as CTO and Chief Scientist, leading its AI research and engineering through its acquisition by Cohere
- ๐ก๏ธ Worked on Large Language Models (LLMs), security, adversarial attacks and interpretability at Google DeepMind
- โ๏ธ Led the Large Language Model team at Stability AI
- ๐ Worked on the Claude model and AI research on the Alignment team at Anthropic
- ๐ Researched scaling, bridging theory and practice, and out-of-distribution robustness as a Research Scientist Intern at DeepMind
- ๐ง Was a Google AI Resident and started and led 5 projects at Google Brain
And a few more academic things about me
- ๐ Got my PhD from Stanford University with Prof Surya Ganguli, was an Open Phil AI Fellow and my work currently has 25k+ citations on Google Scholar
- ๐ Studied physics and mathematics at Trinity College, University of Cambridge (with first-class honours every year)
- ๐ Won (ranked #1 in the world) the International Olympiad on Astronomy and Astrophysics (IOAA) in 2011 (and ranked #2 in 2012 and #8 in 2010), and got two silver medals at the International Physics Olympiad (IPhO)
Miscellaneous
- ๐ Have an academic blog on AI, math, and physics
- ๐ชถ Write an AI research, progress & policy newsletter โจ Pokrok v AI โจ in Czech with Kristina Fort
- ๐ Selected a (Czech) Forbes 30 under 30 in 2022
- ๐๏ธ Advised the Czech President and Minister for Research and Innovation on AI
- ๐จ๐ผโ๐ซ Mentor high-school students for Yoda Mentorship, teach at Discover Summer Academy and co-organize the Astronomy Olympiad
