You’ve probably seen the sleek Uber Freight trucks or at least heard how they’ve basically turned the "Wild West" of long-haul trucking into something resembling a functioning app. But when people talk about the brains behind that operation, they usually point to the logistics veterans. Honestly, they’re missing half the story. To understand how Uber Freight actually works—not just the business side, but the "how is this even possible" side—you have to look at Mike Del Balso.
Mike wasn’t a trucker. He wasn't a freight broker. He was a product manager with a massive problem: how do you make thousands of complex AI decisions every second without the whole system crashing?
The Michelangelo Era: Where it All Started
Before he founded Tecton, Mike Del Balso was the PM lead for Michelangelo. If you aren't a data nerd, that name might not mean much, but inside Uber, it was the "holy grail." Michelangelo was the internal machine learning platform that powered everything. We’re talking about the ETAs you see on the app, the surge pricing that kicks in when it rains, and eventually, the routing for Uber Freight.
The reality of 2015 was messy. Uber had data. Mountains of it. But they couldn't use it fast enough. Mike saw that data scientists were spending like 80% of their time just cleaning data instead of actually building models. It was a massive bottleneck.
Under Mike’s leadership, Michelangelo grew from a small internal tool to a platform supporting tens of thousands of models in production. This wasn't just "regular" ML; this was Operational ML. It means the models weren't just sitting in a lab making predictions for next month—they were making decisions for real people, in real-time, on the road.
Why Mike Del Balso Uber Freight Matters
So, what does a "feature store" guy have to do with moving a 53-foot trailer across the country? Everything.
Logistics is a game of variables. You have traffic, weather, driver fatigue, fuel prices, and warehouse wait times. If you want to automate freight, you can’t just guess. You need a system that can look at a driver in Ohio and a load in Indiana and decide—instantly—if that’s a profitable match.
The Technology Gap
Most people think Uber Freight is just a "matching" app. It’s actually a massive forecasting engine. Mike Del Balso’s work on Michelangelo provided the literal foundation for this.
- Dynamic Pricing: Uber Freight needs to tell a shipper a price right now, even if the truck won't move for three days. That takes predictive modeling at a scale most companies can't touch.
- Feature Engineering: This is Mike's bread and butter. A "feature" is just a piece of data—like "average wait time at this specific warehouse on a Tuesday." Mike’s systems made it so those features were available to the Freight team instantly.
- Reliability: In freight, if the system goes down for ten minutes, you lose millions. Mike’s focus was on "production-ready" AI. Not experimental. Not "kinda" working. But 99.99% uptime.
From Uber to Tecton: The "Feature Store" Revolution
Eventually, Mike realized that the problems he solved at Uber weren't unique to Uber. Every company trying to use AI was hitting the same wall. They had the data, they had the models, but they couldn't connect the two in production.
👉 See also: Why The Tale is Basically What’s Powering Your Google Discover Feed Right Now
He left to co-found Tecton with Jeremy Hermann and Kevin Stumpf. They basically took the "brain" of Michelangelo and turned it into a product the rest of the world could buy. Today, companies like Coinbase and Fortune 500 retailers use Tecton to do exactly what Mike did for Uber Freight: make AI work in the real world.
It's funny. People talk about "AI" like it's some magical cloud, but Mike’s career proves it’s actually a plumbing problem. If the pipes (the data) are clogged, the house (the app) doesn't work.
What Shippers and Carriers Can Learn
If you're in the logistics space, the "Mike Del Balso" approach is basically a blueprint for the future. You can't just buy an "AI tool" and expect it to fix your supply chain. You have to build the infrastructure first.
Actionable Insights for the Future of Logistics:
- Stop ignoring your "features": Don't just collect data. Figure out which specific signals (like weather or dock speed) actually impact your bottom line and make those signals accessible to your team.
- Focus on "Operational" over "Analytical": It's great to know why you lost money last quarter. It's better to have a model that prevents you from losing money on a load right now.
- Modularize your tech: Mike often talks about making AI components "plug-and-play." Don't build one giant, rigid system. Build small, reliable parts that can be swapped out as technology evolves.
- Invest in MLOps: If your data scientists are still manually moving CSV files to "train" models, you aren't doing AI; you're doing expensive paperwork.
The story of Mike Del Balso and Uber Freight isn't just about a successful startup or a clever app. It's about the shift from "guessing" to "knowing." When you see a truck booked through an app in seconds, you aren't just seeing a business transaction. You're seeing the result of years of invisible engineering that started in a small office at Uber.