About
I grew up in San Francisco but have spent most of my life in France and Switzerland. I love working on interdisciplinary problems. Most of my work takes place at the intersection of theory and implementation. I care about understanding why something works, not just that it does.
I compulsively obsess over the long arc of the future. A hundred years from now, I see humanity among the stars having explored beyond Earth and thriving across the solar system. I believe aligned AI will be the key driver of breakthroughs in biology, medicine, materials science, and robotics, and that energy abundance, unlocked by nuclear fusion and solar power in space, will be the foundation for all of it.
What I think about
Artificial intelligence
The economic impact of self-improving and intelligent systems is hard to overstate. At some threshold, capability compounds faster than we can track. I care most about what happens after the LLM scaling era: genuinely autonomous systems, internal representations, and making sure that transition happens in the open and in a politically neutral way. I am particularly interested in latent world models.
Robotics and manufacturing
I spent two years building humanoid software and came out of it convinced that humanoids aren't what will reindustrialize Europe and the US. A general-purpose body is rarely the right tool for a factory. I'm interested in what is: automation that is built around the product, cheap enough for small batches, and fast to set up. Longer term, the same machinery is how we'll build things off-world.
Brain-computer interfaces
The most inherently interdisciplinary field I know. Augmenting cognition and building a safe interface with superintelligent systems is crucial for a smooth transition to AGI. A genuine game-changer for people with disabilities and potentially for all of us. Intro reading.
Decentralized systems
I have experience in smart contracts and high-performance orderbook infrastructure. I believe a decentralized financial system native to the internet is the future of money, and it's encouraging to see traditional finance finally catching up. Anyone should have access to a safe and efficient financial system.
Work and research
Microsoft Research, Student Researcher
2026 to present
Latent world models for robot manipulation on Franka arms in Prof. Marc Pollefeys' Spatial AI Lab. Currently working on latent action prediction for VLAs, building on LaWAM, and extending it to moving egocentric cameras. Also exploring temporal-distance scoring for planning in latent space.
World and video-action models for humanoid control
2026 to present
Training world and video-action models for the Unitree G1 on bimanual data: NVIDIA Cosmos 2.5 paired with an inverse dynamics model, and a Cosmos 3 model that jointly predicts video and actions in the SONIC action space, using human ego-view priors. The Cosmos 3 model runs in real time with NVIDIA's SONIC controller at 50 Hz on the G1.
ETH Robotics Club, Humanoid Software Lead
2024 to 2026
Led AI and software development for the club's humanoid on Unitree G1s: VLA fine-tuning and deployment (NVIDIA GR00T), whole-body control (SONIC), VR teleoperation and data collection pipelines, and industrial pilots with sponsors.
EPFL Machine Learning and Optimization Lab, Research
2022
Research on semantic uncertainty to improve LLM decoding for reasoning tasks under Prof. Martin Jäggi. [paper]
DeFi, smart contracts and infrastructure
2023 to 2024
Built and contributed to multiple high-performance onchain financial protocols on Ethereum, Sui and Solana, including DeepBook, Gelato and Suilend.
Founderful, Student VC Partner
2023
Sourced and led a pre-seed deal for Stacksync (YC F23) at Switzerland's largest pre-seed fund, targeting EPFL and ETH Zurich.
Honey Labs, Co-founder
2020 to 2022
Built one of the first and largest cross-chain money markets on Solana and Ethereum. Seed-funded, acquired 2023.
University of Bern, Astrophysics
2019
Placed 2nd in the Swiss National Science and Youth Competition. Published on gravitational lensing and dark matter distribution in the Milky Way.
Writing
Reading list
- How to do great work, Paul Graham
- The decade ahead, Leopold Aschenbrenner
- The ethos of the divine age, Pirate Wires
- Centauri Dreams, Paul Gilster
- Machines of loving grace, Dario Amodei
- Hackers and painters, Paul Graham
Culture
I like swimming and the gym. Kitesurfing and old sports cars. Here are some artists I enjoy listening to:
- Black Sabbath
- Led Zeppelin
- The Doors
- The Beatles
- The Rolling Stones
- Jimi Hendrix
- AC/DC
Contact
Academic: dfoodeei@ethz.ch
Everything else: darius.foodei@gmail.com