Small Big Things
Small Big Things is a physical AI company. We help robotics and physical AI foundation model teams build better models and deploy them faster.
Models train at cloud speed, but testing them on real robots still moves at lab speed. We want physical experimentation to work like requesting compute: describe the experiment, and we schedule it on real hardware, coordinate the physical resets, safety, and operations, and send the data back.
We're starting with end-effectors, the small part of the robot where a policy meets the world, and where most policies fail.
What we can help with
- Evaluation
- Benchmark your checkpoints on controlled, repeatable tasks.
- Testing
- Stress-test policies on edge cases before they ship.
- Reinforcement learning
- Run real-world rollouts with automated resets.
- Data collection
- Collect on-robot episodes with vision, proprioception, and touch.
Work with us
We're working with a small number of early partners. If physical experimentation is slowing your research down, write to us at hello@smallbigthings.ai.