Cases
A few engagements, written up NDA-safe, names and figures adjusted, the thinking and the outcomes intact.
Building an R&D org and process around AI agents at a connected-hardware company
I built the R&D from scratch, hired the team and set up how specs get written, how code gets reviewed and how releases get verified, and now 5 to 7 engineers own the whole product lifecycle without the other traditional SDLC roles.
Rebuilding an R&D org and a country-scale real-time news platform
As CTO of a national media group I hired full-stack, mobile, QA and DevOps teams from scratch, modernized the stack to microservices on Kubernetes, and led a full rewrite of the real-time news platform so it holds 1 to 2 million users at peak instead of going down at 200 thousand, with 4× lower infrastructure spend.
Standing up national-priority projects in the middle of COVID
Running R&D at a project-based delivery company through COVID, I stood up national-priority teams on immovable deadlines, mobilizing an airport COVID-testing team in four days and shipping a computer-vision rapid-test reader at about 97% accuracy, while running 30+ engineers across parallel programs.
Shipping a logistics MVP with one engineer and a team of AI agents
MyTruckAI is my own freight-logistics platform and the extreme test of the operating model I sell: one engineer plus a team of AI agents, no hires. Agents write the code against specs, run the tests and carry deploys to production while I hold product, architecture and review — so the MVP reached working mobile apps and a backend with near-zero burn at the validation stage.
An AI product built by an AI-run delivery process
Pebble (pebblearn.ai) is AI on both sides: the product takes routine manual work off learning-diagnostics specialists, and it's built by an AI-run delivery process where I'm the only human — co-founder carrying the whole technical side. Every feature goes spec → agent-written code → automated verification → deploy, my review on intent, with near-zero burn on startup cloud credits.
Most of my work can't be named, so these are anonymized.
Client names, exact figures, and identifying details are adjusted or withheld. The architecture decisions, the approach, and the shape of the outcomes are real. Happy to walk through specifics under mutual NDA on a call.