Ben Schmidt

Mastery, not attendance. I built HeyLoopy because I was tired of watching brilliant people get left out to pasture. In my last companies I saw how traditional training fails: we hire great people, hand them static docs and a "good luck," then act surprised when they don't excel. That's not a talent problem. It's a support problem. So we built the opposite of sink-or-swim: daily 60-second reps tied to your own policies and SOPs, so people actually retain what they were trained on and can do the work when it counts. No magic. Just practice, made a lot less painful.

I led the integration of RoadBotics into Michelin, with teams across the US and France. Folding a startup into a storied institution is its own master class in change: different speeds, different stakes, the same people problem underneath. My job was keeping the product, the team, and the customers steady through it.

Co-founded and ran RoadBotics, an AI company that turned street-level imagery into objective infrastructure decisions for cities and engineering firms. Built it, raised for it, and sold it to Michelin in 2022, then stayed on to land the integration without breaking the product, the team, or the customers.

The early days. We used computer vision and machine learning to turn ordinary road imagery into condition assessments a city engineer could trust, so nobody had to do windshield surveys. AI doing the tedious looking so people could make the call. Technology that makes people more capable instead of replacing them, the idea that's followed me ever since.

Ran product and engineering, including the data science and ML teams, back when "machine learning in production" was still a novelty (2013). We shipped real ML into energy markets, and I helped raise the money to keep doing it.

Forecasted price and supply/demand in the PJM electricity market, then turned the models into trading strategies (statistical arbitrage, hedging). Python and large-scale compute, hands on the data. This is where the AI-builder part started.

For my dissertation I studied how networks in the brain communicate. We recorded brain activity non-invasively while people did tasks, then worked out how different regions coordinate, and at which frequencies they "talk" to each other. The honest version: it's a lot of time-series and spectral math on very noisy signals. The patterns were too tangled to eyeball, so we trained machine-learning models to find and predict them. That's where the AI half of my career actually started, on brain data, years before I built it into companies. Five years in that lab left me with a conviction that still runs everything I do: the brain is gloriously messy, and the real answers are almost always less magical than the headlines promise.