Bespoke Labs is an applied AI research and infrastructure company focused on making AI agents more reliable. It builds reinforcement learning environments, data-curation systems and optimization tools that help frontier AI labs and enterprises train, evaluate and improve long-horizon agents. Its research and open-source work includes OpenThoughts, Terminal-Bench, Curator, GEPA, AutoResearchExam and specialized models such as Bespoke-MiniCheck and Bespoke-MiniChart. The company works on realistic digital environments where agents can practice complex tasks, receive measurable feedback and improve before production deployment.
Bespoke Labs offers candidates the opportunity to work on research and production infrastructure at the frontier of reinforcement learning, agent evaluation and AI post-training. The company raised $40 million across seed and Series A financing from investors including Wing Venture Capital, 8VC, Mayfield and The House Fund. Its work includes widely visible open research projects such as OpenThoughts, Terminal-Bench, Curator and GEPA, and collaborations involving companies such as NVIDIA and Intuit. Current engineering roles highlight high ownership, direct work with AI researchers, modern Python and TypeScript systems, competitive compensation and opportunities in both Mountain View and Bangalore.
Backend and AI
Frontend
Frontend
Database
Search and Data
Cloud Infrastructure
Cloud Infrastructure
Co-founder and CEO of Bespoke Labs. Sathiamoorthy previously worked as a Staff Software Engineer at Google DeepMind and Google Brain, including work on large-scale recommendation systems. He earned a B.Tech from IIT Kharagpur and an MS and PhD from the University of Southern California. He started Bespoke Labs in 2024 to advance data curation and reinforcement learning environments.
Co-founder and Chief Science Officer of Bespoke Labs and Professor of EECS at UC Berkeley. Dimakis researches generative AI, information theory and machine learning and has published more than 150 papers. He earned his PhD from UC Berkeley and a Diploma in Electrical and Computer Engineering from the National Technical University of Athens.
Bespoke Labs
Bespoke Labs published research showing how long-horizon machine-learning engineering tasks and self-context distillation were used to post-train NVIDIA's Nemotron 3.5 Super-VL. Its strongest reported model improved its AutoResearchExam score by 24.8% relative to the base model under the tested setup.
Read articleBespoke Labs
Bespoke Labs published work with Intuit and Credit Karma on post-training a specialized small language model, demonstrating its approach to data curation and model specialization for enterprise use cases.
Read articleBespoke Labs
Bespoke Labs announced $40 million raised across its seed and Series A financings. Wing Venture Capital led the Series A and 8VC led the seed round, with participation from Mayfield, The House Fund and prominent technology executives and researchers.
Read articleBespokelabsLocation not specified
BespokelabsMountain View, USA
BespokelabsLocation not specified
BespokelabsLocation not specified
