Summary
Dan Valentine is a full-stack engineer with 11 years of experience who has transitioned from building scalable products at startups to research-focused work on AI safety and interpretability. Based in San Francisco, he has shipped cross-stack systems at companies like Prodigy Education and BioRender and contributed to cybersecurity automation at edgescan, then pivoted to ML safety research with projects at AI Safety Camp, ML Alignment & Theory Scholars, and Anthropic. His research contributions include an ICML 2024 Best Paper on LLM Debate and an ICLR 2025-accepted study on image jailbreak transferability in vision-language models, reflecting a rare blend of production engineering and peer-reviewed research. Studying machine learning full-time, he now works as a Member of Technical Staff at Anthropic, combining hands-on implementation skills with a focus on scalable oversight and model robustness. An applied physics background underpins his analytical approach and explains his appetite for probing model internals as well as building reliable software.
11 years of coding experience
6 years of employment as a software developer
Applied Physics, Applied Physics at Dublin City University