Fais Mehrban · Robotics · Physical AI
I build robot intelligence.
Software that lets robots get real work done, not only see and move. Before this, I spent my career getting hard technology into real use, from fintech infrastructure used by millions to a deep-tech lab and an exoskeleton capturing human motion data on real sites.
Now Building a robot-intelligence company, in stealth, with a team from Stanford, MIT, Princeton, Penn, ETH Zürich, TU Darmstadt and UCL. More soon.
Writing
02One Model Won’t Solve Physical AI
Why robots need to think about their own thinking, not just more data.
Read the essayRobots Need More than VLA and World Models
Position paper, arXiv 2606.06556. Co-authored with leading robotics researchers in the US and Europe.
About
I’ve spent my career taking hard technology and making it work for real users, at scale.
I started in engineering at Cambridge, then moved to law and business at Warwick. As a product owner at Tier 1 financial institutions, I led multinational teams building fintech and regtech infrastructure used by millions. I built and sold fleet management software, then co-founded and led a deep-tech lab in verifiable AI, data oracles and confidential compute, which grew to 160+ people and products reaching 200,000 users.
Robotics is where that problem is hardest. I built an exoskeleton to capture human work, supplied early training data to labs, and set up partnerships with access to more than 10,000 people across real sites. Those labs taught me that data alone doesn’t scale. The bigger gap is intelligence, and that is what I work on now.
The road here
- Now
Robot intelligence
Building, in stealth, the intelligence that lets robots get real work done.
- Robotics
Exoskeleton and human motion data
Built a full-body exoskeleton that captures vision, touch and motion in a single pass. We supplied early training data to labs, directly and through a partner, under strict NDA, and set up partnerships across real sites globally. Alongside it, I co-founded a UK robotics club to bring together people building in the field.
The labs using the data showed us that data provision has limited scale and that the bigger problem was intelligence, so that is where I went next.
Early concept 
The lab - Deep tech
Co-founder & CEO, deep-tech lab
Verifiable AI · data oracles · confidential compute
Built and scaled a deep-tech lab to 160+ people, with a partner ecosystem of 80+ enterprises and products reaching 200,000 users.
- Banking
Product owner, Tier 1 financial institutions
Fintech · regtech
Led large multinational teams that built fintech and regtech apps and infrastructure used by millions of people across banking and tech.
- Built & sold
Fleet management software
Built and sold fleet management and logistics software for rental and last-mile delivery operators.
- M&A
M&A and capital markets
Slaughter and May · J.P. Morgan
Worked on M&A and capital markets at Slaughter and May and J.P. Morgan.
- Early days
Studies, and the first businesses
CambridgeEngineeringWarwickLaw & businessWarwickLLMOxfordAI for BusinessI left an MBA early to build. Along the way I built and sold a hire company, imported and modified Japanese performance cars (Skylines, Supras, Evos and more), and raced on track.
Get in touch