Smiling Buddha is an open research initiative building the real-world data layer for physical AI.
The next decade of robotics and embodied AI will depend on shared data, shared benchmarks, and shared reproducibility standards. Web-scale pipelines built for LLMs don't transfer cleanly to physical intelligence. Smiling Buddha exists to build that missing infrastructure — openly, reproducibly, and with the broader research community as the primary user.
What Hugging Face is for language models, ROS for robotics middleware, and PyTorch for deep learning — Smiling Buddha aims to be for physical-AI training data, benchmarks, and reproducibility. Not a company. Not a platform. A public research substrate anyone can build on.
Every dataset ships with a schema, a validation score, and a reproducibility manifest. Every benchmark ships with data, protocol, reference code, and a public leaderboard. Every paper ships with LaTeX source and the exact data used to produce every figure. Every decision that shapes the project goes through a public RFC. That's the deal.
Smiling Buddha is a community initiative with institutional research support from Pragma One and contributing labs. The initiative is not a commercial entity, does not sell datasets or benchmarks, and does not offer commercial services. Institutional partners underwrite the infrastructure and dataset validation work; contributions are governed by the community RFC process.
Research collaboration, dataset contribution, or institutional partnership questions: [email protected]. Technical questions and bug reports go through GitHub. Real-time discussion happens in Discord (linked from the Community page).
