I do live robot demos, write hands-on guides, and publish open source around NVIDIA's simulation and robot learning stack (Isaac Sim, Isaac Lab, Jetson). Some things I built in this role:

• Physical AI demo in production (Japan). I designed and built the whole thing: a smart factory simulated with Newton and MuJoCo-Warp GPU physics, a Unitree G1 humanoid walking on a reinforcement learning locomotion policy, and a Franka arm. You control all of it with natural language, in English, Spanish or Japanese. It's deployed at an NVIDIA center in Japan with automatic failover, and I trained the local team that operates it today.
• VR teleoperation for the GR00T N1.7 + LeRobot launch with Hugging Face. I built teleoperation of the SO-101 arm in Isaac Lab with a Quest 3 over CloudXR. The streaming was unstable, so I found the root cause and sent two pull requests upstream to NVIDIA Isaac Teleop: PR #726 (frame-rate lifecycle: p95 one-way latency from ~54 ms to 10-13 ms, packet loss from 3.9% to 0.06%) and PR #753 (streaming defaults documentation) (Seeed coverage ↗).
• NemoCast, Computex 2026 (Taipei). A live AI news show hosted by two UBTECH Alpha Mini humanoid robots. Gemma 4, Kokoro TTS and live English to Mandarin translation running locally on a single 8 GB Jetson Orin Nano Super at the NVIDIA booth.
• World model demos (NVIDIA Cosmos). Interactive video generation on 8x H100s: a slider to compare real vs generated footage, and a live viewer that shows every denoising step while the model generates.
• On-device video generation (GTC Taipei). A full text-to-video pipeline (LTX-2.3, 22B parameters) running entirely on one DGX Spark (GB10 Grace Blackwell), no cloud.
• The macOS flashing workflow for DGX Spark. I built it because I needed it for my own units, and NVIDIA adopted it as the official workflow.
• Google Tokyo & Paris (2024-2025). I presented the international debut of Gemma 2 at the Gemma Developer Day in Tokyo, with the demo live on Jetson and a surprise appearance by Google's CEO (video ↗), and the Gemma 3 launch at Google's Paris office.

• Every time a new wave of generative models dropped (GPT, Llama, Mixtral, Qwen, LLaVa, InternVL and other multimodal models), I got them running on Jetson embedded devices as soon as the weights were out, and published the guides and demos for the developer community.
• I ran the Jetson education funnel in EMEA: professors got hardware grants, became Jetson AI Ambassadors, and taught the certified course to their students.
• My bar for every launch demo: a developer should be able to deploy it in under 15 minutes. That meant preconfigured images with samples, shared repos, tutorials, and pull requests.
• I gave the talks and ran the demos at international events, and wrote educational material for partners, universities and enterprise developers.

• I built and led a team of designers, developers and creative technologists to grow the Qiskit developer community. We ran demos at the IBM booth at CES and at the World Economic Forum in Davos, and hackathons in the Swiss Alps and in the mountains outside Tokyo.
• I got into IBM by winning the IBM Quantum hackathon (Qiskit Camp 2019) with an XR quantum circuit viewer.

• I built the world's first 6DOF "Mario Kart" style VR experience in my apartment, on a Lenovo Mirage Solo (video ↗). I presented it at Oculus Connect in San Jose, and it was recognized among the top VR projects of the Launchpad cohort.

• Designed and taught Ethereum smart contract training (Python) for executive and engineering teams in Zurich and Singapore.

• I founded the Technology Lab and designed the "WOW Room", an immersive classroom that won the Reimagine Education Award. Teacher Excellence Award in 2017 and 2018.

• ProUD Alumni: picked as one of the 50 distinguished graduates of the faculty (ProUD ↗).