Engineering notes from the physical world.

Technical writing on robotics datasets, Physical AI infrastructure, and the systems we build along the way.

6 published notes
  1. 5 min read

    Human-centric architectural priors unlock sample-efficient latent spatial reasoning

    Spatial video QA generated from 29 real-world home layouts beats a matched procedural-house baseline by 5.3 points on VSI-Bench, and matches a 25K-example run with 5.3x less data.

  2. 9 min read

    Autoresearch on robotics training data

    An open source recipe for autoresearch on robotics training data.

  3. 2 min read

    HFlow: open-source data quality infrastructure for Physical AI

    HFlow is an open-source SDK for data teams that collect, transform, and curate Physical AI data.

  4. 8 min read

    YouTube System Design for Robotics Data Infrastructure

    What a classic interview question got right about the infrastructure behind robotics data.

  5. 1 min read

    Pareto: high quality data curation for robotics

    Robotics teams need a more rigorous way to understand, curate, and improve training datasets.

  6. 13 min read

    Finetuning PI05 for a custom task, robot and environment

    To understand what it takes to build a humanoid robot with model-based control, we finetuned @physical_int's (PI) Pi05 model for our custom use case and environment. We incurred ~$10K in hardware costs, compared to the typical ~$20K set up (DROID/ALOHA).