Home / Founders / Directory / Tomás Hernando Kofman
Tomás Hernando Kofman
Co-founder and CEO · Not Diamond
Model routing Prompt adaptation
Not Diamond argues that AI will keep fragmenting across models and that the winner will be the layer that routes each request, and Kofman runs it with backing from IBM, SAP.iO Fund, and executives from Google, Hugging Face, and Datadog.
Tomás Hernando Kofman is co-founder and CEO of Not Diamond, a San Francisco company whose model router chooses the best language model for each request and whose Prompt Adaptation product rewrites prompts for new models. He leads it with CTO Tze-Yang Tung and COO Jeffrey Akiki.
Math at Brown, Design at RISD
Kofman studied math at Brown University and design at the Rhode Island School of Design, according to his AI Council speaker profile. The profile says he built a digital civic tech platform that was featured on the cover of TIME, then co-founded a housing startup that he worked on for four years through three rounds of funding.
A Recommendation Layer for Intelligence
Not Diamond’s about page opens with the 1990s portal wars, in which Yahoo, Excite, Lycos, and AOL tried to assemble every service, and Google instead built “a routing algorithm that analyzed the distributed web and sent each query to its best destination.” The company calls itself “the data-driven recommendation layer for intelligence,” and the page lists investors including Jeff Dean, Julien Chaumond, Ion Stoica, Guillermo Rauch, and Olivier Pomel.
Prompt Adaptation and Funding
In May 2025 Not Diamond launched Prompt Adaptation and announced a round with IBM, SAP.iO Fund, Myriad Venture Partners, Deepwater, DNX Ventures, Ambush Capital, and Defy, according to its press release. At AI Council in 2025 Kofman gave a talk titled “How to Build Your Own Model Router,” and the speaker page describes Not Diamond as a multi-model AI infrastructure platform that helps AI/ML teams assess, onboard, and optimize new models.
