Bespoke Labs (https://bespokelabs.ai/), a Mountain View, CA-based applied AI research lab, has raised $40 million across Seed and Series A rounds. The company builds company-scale reinforcement learning environments and infrastructure for training reliable AI agents. The capital will expand its research team and scale production-grade environment curation.
RL Environments Become Agent Infrastructure
The timing aligns with frontier labs shifting focus from models to environments. Mechanize raised roughly $9 million in an angel round in April 2025. Prime Intellect secured $21 million total. HUD raised $16 million Series A. Bespoke Labs differentiates by hiring research scientists to build complex enterprise simulations rather than relying on contractor volume or simple app-level tools.
Agent Failure Rates Highlight Environment Gap
Seventy to 95 percent of AI agent projects fail to reach production. McKinsey data shows 88 percent of organizations use AI in at least one function, yet only about one-third have begun scaling across workflows. Current approaches often rely on static datasets that cannot capture the multi-turn, long-horizon interactions agents encounter in real codebases, Slack threads, and microservices.
Research-First Approach to Environment Building
Bespoke Labs releases open tools including Bespoke Curator for synthetic data curation, the OpenThoughts reasoning dataset with hundreds of thousands of downloads, and Terminal-Bench, which Anthropic, OpenAI, and Google DeepMind use for agent evaluation. Its Genetic Evolutionary Policy Algorithm automates prompt and policy tuning in enterprise settings.
As CEO Mahesh Sathiamoorthy noted:
"Frontier labs, enterprises, and all organizations relying on reliable agents need access to high-quality environments."
Wing VC and 8VC Signal Sector Conviction
Wing Venture Capital led the Series A with participation from Mayfield and The House Fund. 8VC led the earlier Seed round. Angels include Google’s Jeff Dean and executives from Anthropic, OpenAI, and Meta. The backers bring both deep technical expertise and enterprise go-to-market experience.
AI Agents Market Accelerates Toward $53 Billion
The AI agents market is projected to grow from $7.8 billion in 2025 to $52.6 billion by 2030. Multiple forecasts cite 44.9 to 49.6 percent CAGR. Frontier labs are already committing over $1 billion annually to RL environments, positioning specialized infrastructure providers as critical suppliers.
Ex-DeepMind Founders Drive Research Credibility
CEO Mahesh Sathiamoorthy previously worked as a staff research engineer at Google DeepMind. CSO Alex Dimakis is a UC Berkeley professor. The team has released multiple ICLR-accepted papers and maintains an open-source release strategy that has attracted adoption from Meta, Amazon, and Microsoft.
Hiring and Partnerships Signal Execution
The company added 13 team members in June 2026, including a Head of Data Operations from Grammarly and a Data Operations Program Manager from AWS. A recent partnership with Intuit Credit Karma produced a specialized compliance model achieving 99.4 percent accuracy at 15-20x lower cost than commercial APIs.