Bespoke Labs, a San Francisco-based AI infrastructure startup, has announced that it has raised a total of $40 million in funding. The funding consists of two rounds, including an $8.25 million Seed round led by 8VC and a $31.75 million Series A round led by Wing VC.
According to Bespoke Labs, the new funding will be used to expand its research team, scale its environment-building infrastructure, and accelerate its business momentum. The company is focused on building training environments and evaluation systems that help enterprises and AI labs develop more reliable AI agents for real-world use.
About Bespoke Labs
Founded in 2024 by Mahesh Sathiamoorthy, a former Google DeepMind engineer, and Alex Dimakis, a professor at UC Berkeley and the company’s Chief Scientist, Bespoke Labs builds infrastructure for training and evaluating reliable AI agents.
The company develops tools for creating high-quality post-training datasets, realistic training environments, and evaluation systems that help enterprises and frontier AI labs improve the performance of large language models in real-world applications.
What Bespoke Labs Actually Builds
Bespoke Labs develops the infrastructure that helps AI agents learn, improve, and perform reliably in real-world business environments. Instead of focusing on building another AI model, the company creates realistic training environments where AI agents can practice complex tasks, generate high-quality post-training data, and be evaluated before they are deployed.
These environments are designed to mirror how businesses actually operate, giving AI systems a safer and more effective way to learn from experience.
The company also builds tools that support every stage of the post-training process, including data curation, prompt optimization, and model evaluation. Its open-source projects, such as Curator, OpenThoughts, Terminal-Bench, and GEPA, are used by AI researchers and enterprises to create better datasets, benchmark AI agents, and improve model performance.
By combining these tools with realistic training environments, Bespoke Labs aims to make AI agents more reliable for enterprise use cases where accuracy and consistency matter most.
Funding History
The company first raised an $8.25 million Seed round in June 2024, led by 8VC, with participation from Jeff Dean, Resolve AI CEO Spiros Xanthos, DevRev CEO Dheeraj Pandey, and other investors.
On 6 July 2026, Bespoke Labs announced its $31.75 million Series A round led by Wing VC, with participation from Mayfield, The House Fund, dbt Labs CEO Tristan Handy, and angel investors from Anthropic, OpenAI, and Meta. With this latest investment, the company has raised a total of $40 million to date.
“The next frontier in AI is not just better models, but better environments for training and evaluating reliable AI agents. Bespoke Labs is building that critical infrastructure.” — Navin Chaddha, Managing Partner, Mayfield
“For the past two years, we’ve focused on advancing data curation research and building reinforcement learning environments for AI agents. This funding will help us expand both efforts.” — Bespoke Labs Team
The Next Phase of AI Infrastructure
As AI adoption expands across industries, investors are increasingly looking beyond applications to the infrastructure that makes AI systems more reliable and production-ready.
As more businesses deploy AI agents for specialized workflows, companies building the tools to train, evaluate, and improve those agents are likely to play an increasingly important role.
In many ways, Bespoke Labs is doing for AI agents what driving simulators did for autonomous vehicles by creating realistic environments where systems can learn before operating in the real world. With its latest funding, the company is well positioned to accelerate that vision and help shape the next generation of reliable AI agents.
