Physical AI software · early

Macrodata Labs

Corporate marketdata-engineCurated record

macrodata.co

Macrodata Labs builds data-processing infrastructure for physical-AI training, founded by Guilherme Penedo and Hynek Kydlicek, the Hugging Face researchers behind the FineWeb and FinePDFs open LLM pre-training datasets. Its first product, Refiner, is an open-source Python framework that ingests the formats robotics teams actually use (LeRobot, HDF5/ALOHA/robomimic/LIBERO, Zarr, MCAP, raw video, Hugging Face datasets) and turns raw robot episodes into training-ready datasets. The company raised a $4M pre-seed led by Air Street Capital in June 2026.