Bagel

Bagel is a decentralized machine learning data lake designed to address the challenges of data sharing and contributor incentives in open-source AI. Its core differentiator lies in combining cryptographic techniques with privacy-preserving machine learning, enabling data and model contributions to be verified and rewarded without exposing underlying proprietary information. For example, its ZKLoRA technology allows developers to prove the effectiveness of fine-tuned AI adapters without revealing the adapter weights, solving the trust and attribution problem in model sharing.

The market pain point Bagel addresses is the 'free-rider' problem in open-source AI: contributors of data or compute often receive no recognition or compensation, which hinders the growth of high-quality open datasets. By introducing cryptographic attribution and incentive mechanisms, Bagel aims to make data contribution a verifiable and economically sustainable activity.

In the past six months (2026-02-28 to 2026-08-31), Bagel has shown active community engagement and technical progress. It hosted a robotics and physical AI community event in Toronto on August 27, 2026, and previously co-hosted a gathering on world models during NVIDIA GTC week in Menlo Park. In March 2026, a Bagel Labs paper was accepted at CVPR 2026, indicating ongoing research contributions to diffusion model development. These activities reflect a focus on building a community around decentralized AI infrastructure rather than purely commercial promotion.