deepmind-alumni
(3)
Phaidra
Phaidra is a Seattle-based industrial AI company that builds self-learning, reinforcement-learning control agents for the world's most complex mission-critical facilities - chief among them the data centers powering the AI boom. Founded in 2019 by alumni of Google DeepMind and the HVAC and defense industries, Phaidra's software plugs into existing building automation and SCADA systems without new hardware, learns a facility's behavior, and continuously adjusts cooling, power and workload setpoints to cut energy use while improving reliability. The company raised a $50M+ Series B in October 2025 led by Collaborative Fund with participation from NVIDIA, positioning itself to make 'AI factories' radically more resource-efficient.
Reflection AI
Reflection AI is a New York-based artificial intelligence lab founded in 2024 by former Google DeepMind researchers Misha Laskin and Ioannis Antonoglou. It is building autonomous coding agents and open frontier language models, with the stated ambition of reaching superintelligence by first solving software engineering. Its flagship product, Asimov, is a code research and comprehension agent for large codebases. After raising $2 billion at an $8 billion valuation in October 2025, the company has positioned itself as an open, Western alternative to closed labs like OpenAI and Anthropic and to Chinese open-weight models like DeepSeek.
Jeremy Fraenkel
Jeremy Fraenkel is the CEO and co-founder of Fundamental, the AI lab that emerged from stealth in February 2026 as a $1.4 billion unicorn with $255 million in funding. A DeepMind alumnus with a graduate degree in Machine Learning from UC Berkeley, Fraenkel previously cut his teeth at JPMorgan and Bridgewater Associates before founding and exiting two startups. At Fundamental, he and co-founders Marta Garnelo (Chief Science Officer, ex-DeepMind/Isomorphic Labs) and Gabriel Suissa are building NEXUS, a Large Tabular Model that does for structured enterprise data what LLMs did for language - delivering accurate predictions from raw tables with a single line of code, no preprocessing required.