Bespoke Labs, a startup specializing in the post-training stage of AI development, announced a $40 million funding round led by Wing VC. The capital will support enhancements to its reinforcement learning platform and further AI data research.
- Raised $40M in two funding tranches including a $31.75M Series A led by Wing VC
- Platform automates creation of reinforcement learning environments with expert input
- Open-source contributions include prompt engineering and a large supervised fine-tuning dataset
What happened
Bespoke Labs Inc. announced it has raised a total of $40 million in capital through two funding rounds. The larger share was a $31.75 million Series A round led by Wing VC and included participation from Mayfield, The House Fund, and employees from major tech companies like Anthropic PBC. An earlier tranche of $8.25 million came from investors including Jeff Dean, chief scientist at Google DeepMind.
The startup focuses on improving the post-training phase of AI development, specifically enhancing reinforcement learning workflows. Their platform automates the creation of simulated training environments tailored for different AI projects, enabling faster and more effective fine-tuning of models.
Why it matters
Post-training, which includes reinforcement learning and supervised fine-tuning, is essential for sharpening AI model capabilities such as reasoning and completing long-term tasks. However, these processes are often complex and time-consuming because they require extensive environment setup, sample task design, and iterative tuning.
Bespoke Labs streamlines these challenges by leveraging automation and contributions from a network of human experts to generate rich, custom training environments. The company also offers open-source tools including GEPA for automating prompt engineering and datasets like OpenThoughts to facilitate supervised fine-tuning, setting a new standard for efficiency and output quality in AI training.
What to watch next
Bespoke Labs plans to deploy the new funding to expand and enhance its reinforcement learning platform. This includes improving its sandboxing technology to reduce latency and increase throughput during AI model training sessions.
Additionally, the company will invest further into AI data research to refine post-training methodologies and continue growing its open-source initiatives. These efforts could significantly influence how AI teams build and optimize models across various industries.