A&S-led startup CentML is making machine learning accessible to everyone

May 15, 2024 by Coby Zucker - A&S News

With its strong ties to the Faculty of Arts & Science’s Department of Computer Science, CentML is making accessible, cost-effective and high-performance machine learning available to companies of all shapes and sizes.

Founded by a team including faculty members, PhD students and alumni, the company sprung from research going on at CEO Gennady Pekhimenko’s lab and the entrepreneurial acumen of his friend and COO, Akbar Nurlybayev.

“I’m incredibly proud of what we’ve accomplished at CentML and grateful for the support from the Department of Computer Science at U of T,” says Pekhimenko, a faculty member and alum who earned his master of science in 2008. “Our success is a collective achievement, reflecting our team’s dedication and the rich academic environment at U of T that fosters innovation.”

As the name implies, CentML provides cost-effective machine learning solutions to customers. At a time when top-of-the-line graphics processing units (GPUs) are being snapped up by big companies for their own AI ventures, CentML has figured out how to get better performance out of cost-effective GPUs. This enables CentML to give companies the best bang for their buck when hosting their machine learning models.

Late last year, CentML secured $37 million in funding from Google, NVIDIA, Radical Ventures, Deloitte and Thomson Reuters Ventures, among other big investors. It’s a massive validation of everything CentML and the small team behind it have done over the past two years.

“It’s been incredibly fulfilling to see CentML’s growth and the recognition of our efforts through successful funding rounds,” Pekhimenko says.

CentML’s association with U of T and computer science extends beyond its leadership. As the company scales up, Pekhimenko and Nurlybayev increasingly bring on students and alumni of the program.

“These individuals are not only technically proficient but also bring a problem-solving approach that is crucial for a startup like CentML,” Pekhimenko says.

“The computer science degree itself gives you a foundation you build on,” says Nurlybayev, who earned his honours bachelor of science in 2007 as a member of New College. “The technologies that are built on top change, but fundamentals stay the same.”

The influx of cash and support CentML has seen comes on the wave of a global increase in interest around AI. According to Nurlybayev, conversations ahead of this year’s funding round were different than when CentML started in 2022. The release of Stable Diffusion and ChatGPT familiarized AI in the public sphere and made generating buzz for CentML’s vision easier.

“Everyone suddenly had a need to reduce their cloud spending or make better use of the very limited chips in the cloud,” Nurlybayev says.

The final two co-founders of CentML are chief architect Anand Jayarajan and CTO Shang Wang, PhD students in Pekhimenko’s lab. The trio of Jayarajan, Wang and Pekhimenko conduct research into optimizing machine learning models and running them on cost-effective hardware — core ideas that underpin CentML’s entire platform.

A group of people from Department of Computer Science, CentML.
The CentML team includes students and alumni from the Department of Computer Science.

According to Pekhimenko, the move into industry was a natural progression for himself and his students.

“Their fresh perspectives and academic rigor were instrumental in shaping our early development,” Pekhimenko says. “The transition was a learning curve for all of us, as we adapted from our previous professional and academic mindset to a more product-driven approach, balancing innovation with market needs.”

But having a good idea is only half the battle. When it came to the task of building a company from the ground up, Pekhimenko turned to his old friend and fellow U of T alum, Nurlybayev, who he met through the Russian-language community at the university.

“Reuniting with Akbar has been fantastic,” Pekhimenko says. “His industry experience, coupled with a deep understanding of machine learning, has brought a unique perspective to our team.”

Together, the team of Pekhimenko, Nurlybayev, Wang and Jayarajan has turned a research subject into a successful company, but the process has been anything but easy. Only recently, with the latest funding round and a move to a new office in October 2023, has everything come together.

“Early on as founders, we had to do a lot of things by ourselves. Now, we are expanding the team with talented and hardworking folks who are capable of sharing various responsibilities with us,” says Wang, who received his master of science in 2020 and his bachelor of applied science from the Faculty of Applied Science & Engineering in 2018. “It's definitely been a journey. Jumping on board has been a learning experience every day.”

CentML’s success validates the belief of Pekhimenko and his team that optimizing machine learning is a crucial next step in AI innovation, economic growth and fostering a more competitive and dynamic market.

“Looking ahead, we’re focused on expanding our product offerings, exploring new markets, and continuing our research and development to stay at the forefront of machine learning technology,” Pekhimenko says. “Bringing the best and most cost-efficient solution to our clients is our everyday mission and we are committed to doing just that.”