Lightning Twice
On Arize’s exit, and twenty years with the founders who built it.
Congratulations to Jason, Aparna and the whole Arize AI team for signing a definitive agreement to be acquired by Dynatrace for $1 billion. I’ve worked with Jason for almost twenty years now across two companies, and this one is special to me. So let me just tell the story.
I met Jason when our companies merged. He and his co-founder Adam joined John and me at TubeMogul when we were only a couple years into our own journey. Few plan to combine two private startups. It’s awkward, egos get involved, cap tables get messy. But honestly, doubling our founder power was one of the reasons we were eventually able to take the company public and sell to Adobe. Remember that part, because it happens again.
I had never worked with anyone like Jason. He was one of the most technical people at the company. With an uncanny ability to see around corners and figure out what to build next, he could sell and dazzle clients with his intellect — and he has an insane engine. We worked non-stop together for eleven years. His job as Chief Strategy Officer, with John running product and Adam running engineering, was to keep us the innovation leader in the space. And oh did he. The list of firsts got ridiculous: first video DSP, first open video viewability standard, first to automate linear TV buying, first after first in brand measurement and safety. It was no fun to compete with us in our prime, and a lot of that was Jason. Underneath the product, we built a culture of doing what you say, figuring shit out, and getting shit done. Hundreds of former TubeMogulers will tell you it was the best run of their careers.
TubeMogul was also an early user of machine learning for media buying and ad optimization, and Jason helped run that team. He read ML research papers on the Peloton, for fun. Once, when our fundraising story needed more oomph, he convinced an ML professor at Columbia and the venture fund the professor worked with to invest in us by out-arguing them on the science. More than once, his team unmasked botnets committing tens of millions of dollars in ad fraud before Google or Microsoft caught them. Not a normal executive.
Get Swift Ventures’s stories in your inbox
Join Medium for free to get updates from this writer.
Getting ML to work at scale back then was brutal. Clients would call to tell us optimization had stopped working in their region before we had any idea anything was wrong. Our telemetry couldn’t see the failures, so we’d throw our very expensive data scientists at the problem and hope. Jason lived inside that pain for years. Most people complain about a problem like that. Jason filed it away as a company.
One more thing about that data science + ML team. It included a promising young ML engineer in her first job out of Berkeley named Aparna. She went on to Apple, then Uber to help build core components of Michaelangelo (Uber’s pioneering large-scale ML infra platform), then started her own company and got into YC. When Jason was getting Arize going, the two joined forces (one of the best intros I’ve ever made). Two founders combining again, the same move that worked for us fifteen years earlier. Aparna became indispensable as a product and company leader and over time the industry’s most recognizable voice on evals and AI observability.
I remember the founding moment exactly. Jason and I were riding mountain bikes up a hot fire road on Mt. Tam sometime after we’d all left Adobe, and he told me he was thinking of starting an ML observability company. It was a napkin sketch. He asked if I’d be his independent director and I said yes on the spot, less because I grasped the full vision than because working with Jason again was the whole draw. Within weeks the napkin was a working prototype and he was out pitching it.
Arize started as an MLOps company, solving the exact problems Jason had lived at TubeMogul. Then generative AI showed up and the industry basically stopped building many traditional ML models. Who would have known, in 2020? There’s a graveyard of MLOps companies that didn’t survive that shift. Jason and Aparna turned the company to monitor and evaluate LLMs and agents without breaking stride. Pulling off one pivot like that is lucky, but doing it twice with the same people isn’t.
The team filled up with TubeMogul stars: Michael Schiff, Gabe Barcelos, Liz Cardenas, Mike Downs, Nathaniel Mar, David Burch, Chris Chayni. People who already knew how to work together and shared the bias for action. Early TubeMogul investors backed them again too, Ashu Garg at Foundation Capital and Ajay Chopra at Trinity. At one point they even hired my brother, Matt Wilson, who’s been their amazing Head of Sales for years now. The mafia runs deep.
And the firsts started stacking up again. At TubeMogul, Jason invented the open viewability standard. At Arize, the team created OpenInference, the first open telemetry standard built specifically for AI, LLM, and agent traces. First open-source eval tool. First AI observability company to bridge pre-production evals with production monitoring. First platform to treat evals, traces, prompts, datasets, and production feedback as one continuous workflow. First continual learning platform for AI agents. Same instinct, a decade apart: figure out what everyone will need in three years, build it first and then open the standard so the market moves toward you. The result is the combination almost nobody gets, real enterprise revenue and real developer mindshare at the same time. Hundreds of enterprise customers, thousands of companies running Phoenix.
As a board member there frankly wasn’t much to fix. My most consistent feedback over the years was “Jason, you’re being too nice,” which is one of many reasons his people love him. Part of why he and I clicked from the start is that we both come from blue collar families with colorful histories. Nobody handed us any of this. Watching what he’s built, twice now, I’m proud of him way beyond the board seat, and I’m so happy for his family.
Eighteen years ago, two sets of founders merged their companies and rode it all the way to Adobe. This week, one of those founders did it again. Arize, the company Jason built with Aparna, signed a definitive agreement to be acquired by Dynatrace for $1 billion. Jason, thank you for letting me be part of the adventure. Dynatrace has no idea how much fun they’re about to have.
