I run AI field tests of tools and agents. Each one gets a real engineering problem, and I check its work against what reality required. The grounding is 15 years across automotive, robotics, and embedded systems, where the gap between plausible output and a working system is expensive.
Two hundred case studies during my EMBA left me with a four-layer lens for the places a system quietly falls out of alignment. These are the same four seams AI is widening right now, and the ones I field-test against.
Layer 01
A CFO signs off on an agent run that burns six figures to return a fraction of that in real value.
Layer 02
Code gets generated faster than anyone can review it or own the way it fails.
Layer 03
A product manager out-ships the whole engineering team on specs by an order of magnitude.
Layer 04
Individual velocity triples while team coordination stays exactly where it was.
AI is very good at producing work that looks finished. Whether it survives contact with real constraints, hardware, timing, safety, the organization around it, is a different question, and answering it takes judgment you only get from shipping the thing yourself. I’m the operator inside these systems, not a consultant reading about them from outside.
So I test it in the open. Every field test gets a real problem, a number, and a verdict, passed, failed, overengineered, whatever it earns. This site is the running record.
I build across the full stack of physical and digital: mechanical, hardware, software, end product. My own lab runs CNC machines, laser cutters, 3D printers, a photography studio, and servers for my own infrastructure.
Personal projects span agritech, photography, and drones. I build my own devices for agricultural experiments at home, build my own robots and drones, and use the studio to document the work.
Over 200 case studies from HBR, Oxford, and INSEAD shaped how I read failure. I’ve published 14 of them here as written breakdowns: Nokia, Boeing, Theranos, and the patterns they share with every deep-tech team shipping today.
2024 - 2025
Took an idea from zero to an operating company with a working delivery model.
2019 - 2024
Helped scale the company from early stage into larger industrial programs and global operations.
2016 - 2019
Managed software programs for major automotive customers including Panasonic, Daimler, and BMW.
2013 - 2016
Built C++ systems used in semiconductor verification workflows at global scale.
2011 - 2013
Led software development work for automotive production programs and embedded control systems.
ESMT Berlin · 2020 - 2022
University of Oxford
FGV EAESP · 2022
Ain Shams University · 2006 - 2011