Meet Scott Clark, an OSU science alum who built a $30M AI startup

How the College of Science prepared him to change the landscape of AI

When Clark arrived as a freshman at Oregon State in 2008, he met Rubin Landau, who ran the (since discontinued) computational physics program at the time. Clark became especially enthralled with Landau's work after checking out one of his books from OSU's Valley Library.

"It was everything that I loved, all put together," Clark said. "It was physics with math on computers. It was like, 'Here's how you teach a computer physics.'"

He convinced his advisor to let him take a 400-level class as a freshman. Although he started out as a math and computer science double major, he switched to a triple major in math, physics and computational physics halfway through his freshman year.

"One of the things I loved about Oregon State was that they let me explore as much as possible."

Oregon State has one of the best numerical analysis departments in the world with legendary professors such as Malgorzata Peszynska and Bob Higdon, he added.

"One of the things I loved about Oregon State was that they let me explore as much as possible," Clark said. "They never told me I wasn't allowed to take a class because I didn't have the prerequisite, or it was a 600-level class, and I was only a sophomore. They always let me try it, to jump into the deep end and see if I could swim."

Clark began expanding his work, pursuing summer Research Experiences for Undergraduates (REUs) in biophysics at the University of California at Davis. He also worked at the Max Planck Institute in Germany, doing computational quantum mechanics work.

He and Malgorzata Peszynska spent a summer working on fluid dynamics problems.

"What I really loved was all this interdisciplinary work of how you fit these fields together and using computational simulations to push that forward," Clark said.

After graduating from Oregon State in 2012, he pursued graduate school at Cornell University in its applied mathematics department. "I could work on all these problems that would otherwise fall through the cracks," he said.

Clark soon received a fellowship from the U.S. Department of Energy, allowing him to work in computational science at various national laboratories, including Los Alamos and Lawrence Berkeley. The latter focuses on scientific discovery and solving large-scale problems in energy, health and the environment.

"A lot of these problems end up looking similar," Clark said. "They end up looking like difficult optimization problems where you build some big, complex machine, system or algorithm, and you want to make it work as well as possible."

Figuring out how to use computers to solve extremely hard optimization problems became his next passion, Companies often spent exorbitant amounts of money grappling to increase optimization. Clark saw an opportunity. "In as few attempts as possible, you want to get to the best possible answer," he said.

Toward the end of graduate school, he faced a choice -- heading to academia or industry. He decided it would be more interesting and fast-paced to at least try industry first.

An internship with Bloomberg in New York City working in finances helped finalize his decision. Clark recalled the thrill of working at a fast pace with highly intelligent people.

"Everyone had something to optimize."

He took the private-sector skills he acquired in New York City and began working for Yelp. The company needed help building a targeting system as it ventured into advertising. "This was like a really fun, hard math problem," Clark recalled. "There was enough scale to make it really interesting, but it was early enough that it was very greenfield, so I wasn't trying to make something slightly better. We were starting from scratch, basically."

It struck him other companies could benefit from what he was doing for Yelp. He created MOE -- the Metric Optimization Engine -- and presented it to companies like Google, Facebook and OpenAI.

"Everyone had something to optimize," Clark said. "I learned this was an extremely pervasive problem."

In 2014, Clark founded SigOpt to market the software he developed to use machine learning and complex algorithms to optimize user experiments for websites and applications. He improved data collection for everyone from federal intelligence agencies to hedge funds and big banks. Traveling the world, he built a team of 30 people and raised $17 million in seven years.

After selling the company to Intel in 2020, he became Intel's vice president of AI and High Performance Computing Software teams within the Supercomputing Group -- helping people build even bigger and better computers. However, after two years, a thought kept nagging him. Customers didn't necessarily want the best result. They wanted the most reliable result.

Thus was born Distributional Inc.

The company creates reliable AI by creating better testing, automating test creations and calibration -- allowing organizations to reduce risk by standardizing testing throughout the life of all their AI technology.

"It's all about how to test them and make sure they're behaving as we wish," Clark said.

A lifetime love of computers

Clark has been working with computers since he was 10 years old in the early 1990s, and his father brought discarded office computers home.

"I tinkered around with QBasic and stuff like that,” he said. “I was always interested in making machines do things that I found interesting, even if it was just printing my name over and over on one of those old terminals."

He built computers for people while he attended Central Catholic High School in Southeast Portland -- customizing early Dell and Gateway computers. "I was willing to value my time at $10 per hour so I could undercut someone who was just updating HTML."

Clark built several dozen computers for local businesses and friends and designed websites. Then his interests took a cosmic turn.

"I became interested in more than how computers work and became interested in how the world and the universe work," he said.

He was particularly interested in physics and began reading Richard Feynman's lectures on theoretical physics. This led to Clark taking advanced placement courses as well as classes at Portland Community College.

"I wanted to understand physics, because that's the way the world works, and I wanted to understand math because that's the language of physics," Clark said.

"I was never told anything was impossible, so I never stopped."

However, his fascination with computers remained. "I was always interested in how you can make computers much faster, how you can run simulations and things like that," he said. "I was enthralled by these very beautiful simulations of fluid dynamics and partial differential equations."

Clark, now 38, said everyone will be touched by AI -- even the engineers who create and service the machines

"I see a lot of companies now not hiring engineers just because it's so efficient to have a system that writes a lot of stuff too," he said. "Engineers are definitely going to be touched by AI. The pace of change is insane. Even working in the field, staying on top of AI is hard. It’s definitely exponential."

Clark's parents weren't engineers or scientists, but while they knew nothing about creating artificial intelligence, he said they knew a lot about nurturing it naturally.

"I'm definitely here as a product of the encouragement of my parents," said Clark, who was born in Salem and raised in Tigard.

"We were exploring together," he said. "I remember my parents were very quickly unable to help me with my math homework anymore, but they were always very encouraging. They'd let me take over the entire dining room table."

It was that kind of encouragement that made him the tech entrepreneur he is today, Clark said. It was the support he also found at Oregon State.

"I was never told anything was impossible, so I never stopped," he said.