 |
 |
 |
 |
 |
I'm a robotics engineer focused on solving real-world deployment problems. Physical AI is advancing fast, but almost none of it reaches a real factory floor. The gap between a model that works in simulation and a robot that works in production is where I work: VLA models, world models, SLAM and navigation. I lead solutions engineering at Neuracore and founded ARTHÉ, an open-source robotics community.
No hype: real deployments, real failures, and what I learned fixing them.
|
Research tooling · LaTeX · AI The research operating system for robotics. Find new papers from arXiv, ask any paper a question and get cited answers, organise your reading list, and write your own in a LaTeX editor with inline AI commands. The hero art on this profile is Tadashi's own home-page robots. |
MediaPipe · RTMPose · computer vision Open-source bike fitting from a 30-second phone video. It tracks your joints through every pedal stroke and returns measured angles with confidence ranges and a ranked plan for saddle, reach and bars. Free, private and fully offline. |
|
Gemini Robotics ER 2.0 · VLA 2.0 · ROS 2 A developer cookbook for Google DeepMind's Gemini Robotics: 31 prompt cards for spatial grounding, grasping, motion planning and safety, a ROS 2 bridge, and the official benchmarks. Translated into 4 languages. |
Hugging Face · PyTorch · robot learning Open pull requests to Hugging Face's LeRobot evaluation pipeline: persisting recorded datasets across batched evals and adding provenance metadata to eval results. Also contributing fixes to VLA-Adapter. |
More projects
| Project | What it does |
|---|---|
| GemmaSense | Hybrid edge-cloud vision-language system: on-device PaliGemma 2 escalating to Gemini when confidence drops |
| Open-ENPIRE-Gripper | Open design of NVIDIA's ENPIRE compliant gripper, adapted to other robot arms (Robotiq Hand-E, OpenArm) |
| WorldLab Robotics Examples | Photorealistic digital-twin environments for robot training, generated with World Labs |
| Reachy-Mini · Gemini + PersonaPlex | An empathetic desktop robot with Gemini vision, voice wake-words and 15+ physical expressions |
| mujoco-playground | Play three robots in the browser: G1 humanoid, bimanual Franka stacking and a Husky on Mars, scored like a benchmark |
| UPenn Robotics Specialization | Aerial robotics, perception, estimation, planning and mobility coursework |
Research collaboration, talks and partnerships, or a mentoring call.
