Carolyn
Mooney

A discipline that didn't have a name

Carolyn Mooney runs Nextmv, a platform that does something deceptively dull: it helps developers build, test, and ship the models that decide which driver gets which order, which worker takes which shift, which box holds which item.

For years, this was where good ideas went to die. A data scientist would build a brilliant optimization model, prove it worked on a laptop, and then watch it stall - because there was no clean way to version it, test it against real traffic, monitor it in production, or hand it to an engineer without a translation layer of heroics. Mooney saw this gap from both sides. She named the fix DecisionOps, and built the tooling to back the word up.

The premise is borrowed openly from software engineering. Developers long ago decided that code should be versioned, tested, reviewed, and deployed through repeatable pipelines. Mooney's argument is that decisions - the algorithmic kind, made thousands of times a second across logistics, scheduling, and pricing - deserve exactly the same rigor. Treat decisions as code. Version them. Shadow-test them against live data. Roll them back when they misbehave. It is an unglamorous proposition, and that is precisely the point.

In May 2026, the bet paid off in the way startup bets are scored: FICO acquired Nextmv, announced on stage at FICO World 2026. FICO - the company most people know only as the keeper of their credit score - is in fact a decades-old leader in decision intelligence. Folding Nextmv into the FICO Platform was a vote that DecisionOps belongs in the enterprise core, not at its experimental edge. Mooney and her co-founder Ryan O'Neil framed the move as a way to push toward agentic decision workflows: systems that don't just recommend a decision, but carry it out and keep a record of why.

In the announcement, Mooney and O'Neil made three promises that read as a tidy summary of what they care about. Stay dedicated to existing customers as they scale up their decision technology. Keep the platform open - compatible with whatever solvers, frameworks, and engines a team already uses, open-source or commercial. And keep building toward agentic decision workflows backed by a comprehensive record of what was decided and why. That last commitment - the audit trail - is the part that turns a clever model into something an enterprise can actually trust.

Missiles, then pizza

Mooney's resume reads like a dare. She studied Systems Engineering at the University of Pennsylvania, earning a BSE, and started her career at Lockheed Martin running large-scale simulations in the ballistic missile and radar world. It is about as far from a delivery app as engineering gets - high-stakes, slow-moving, and obsessed with modeling how complex systems behave under uncertainty.

Then she did something strange with that training. She joined Zoomer, a meal-delivery startup, leading decision engineering - which is to say, she took the simulation discipline of defense work and aimed it at the very fast, very messy problem of getting hot food across a city. From there she moved to Grubhub, where she led the Systems Engineering team and consulted across the business on everything from improving ETAs to scheduling and market management.

That arc is the whole story in miniature. The same systems-thinking toolkit - model the world, simulate the options, optimize the outcome - works whether the payload is a warhead or a pepperoni pie. What changed between Lockheed and Grubhub was the clock. Defense systems give you years. A delivery network gives you seconds. And in those seconds, Mooney kept running into the same wall: the models were good, but shipping them was agony.

The problems she worked on at Grubhub are the kind most people never think about because they only notice when they break. How long until the food arrives? That ETA is an optimization problem. Which courier should take which order, and in what sequence? Routing. How many drivers does a market need on a rainy Friday night? Scheduling under uncertainty. Each one is a decision a computer makes thousands of times an hour, and each one quietly determines whether the food is hot and the driver is paid fairly. Mooney spent years on the inside of that machine, which is why she trusts the unglamorous parts of the problem more than the flashy ones.

^ The DecisionOps loop, roughly. Mooney's pitch: every optimization team should run this cycle as routinely as software teams run CI/CD.

Two operators who'd lived the problem

Mooney co-founded Nextmv with Ryan O'Neil, an optimization specialist who'd worked alongside her at both Zoomer and Grubhub. They were not academics theorizing about a market - they were operators who had personally built the delivery-consolidation algorithms and managed the distributed teams that ran them. They knew exactly which parts hurt.

The company went through Y Combinator in 2020 and, less than a year after launch, closed a $2.7M seed round - a notable raise for a Philadelphia-founded, distributed optimization startup at a time when most venture attention pooled on the coasts. In November 2022 came the $8M Series A, bringing total funding to roughly $19.3M. The money funded a globally distributed team of decision engineers and operations researchers, the kind of people who get genuinely excited about vehicle routing problems.

What makes Mooney's category bet interesting is how early she committed to a word nobody was using. DecisionOps wasn't a market she entered - it was a market she had to first convince people existed. She did it the patient way: talks at INFORMS, fireside chats with investors like Matt Turck, appearances on engineering podcasts, a lecture at Bucknell titled "Decisions as Code," and a community event she named DecisionFest. The acquisition is the clearest signal that the framing stuck.

There is also something worth noting about where she built it. Nextmv is a Philadelphia company, and Mooney kept it that way through a fundraising climate that rewarded relocating to San Francisco. The early seed round was, by the local press's account, a meaningful moment for the city's startup scene - proof that a deep-tech optimization company could raise serious money from a base outside the usual venture corridors. The team grew distributed and global, but the center of gravity stayed in Philadelphia, where Mooney studied and lives.

The Chief Data Officer is a dog

For all the heavy machinery of optimization theory, Mooney keeps a light touch. Nextmv's team page lists a Chief Data Officer named Wally, who is her dog, and whose primary contribution to the company is being at the park. It is a small joke, but a revealing one - a founder who takes the work seriously without taking herself too seriously.

Away from the laptop, she coaches volleyball with Jersey Juniors. There is a neat symmetry there too: coaching is, at bottom, a decision-optimization problem played at human speed - read the system, run the scenarios, make the call, watch what happens, adjust. Mooney has just spent a career doing the same thing with code.

She is, by the evidence, a systems thinker first and a founder second - someone who looked at a recurring frustration across three very different companies and decided the right response wasn't another point solution but a whole discipline. The missiles taught her to model the world. The pizza taught her to do it fast. Nextmv was where she taught everyone else to ship the result.

It is tempting to read the FICO acquisition as the end of the story, but the more accurate reading is the opposite. A category that gets absorbed into one of the oldest names in decision intelligence isn't fading - it's being installed as plumbing. The phrase Mooney spent years explaining to skeptical rooms is now attached to a platform that touches a meaningful slice of the world's lending, fraud, and operational decisions. The unglamorous bet - that decisions deserve the same engineering rigor as code - turned out to be the kind that compounds quietly until everyone assumes it was obvious all along.

Go deeper