By: Erin Vollick
9 Oct, 2023
Gennady Pekhimenko, Vector Institute: AI that teaches itself to be faster and cleaner
October 9, 2023
Gennady Pekhimenko, Canada CIFAR AI Chair, Vector Institute
Machine learning (ML) programs use gigantic amounts of computing – often very inefficiently. Canada CIFAR AI Chair Gennady Pekhimenko’s research solves this inefficiency, making ML models faster and more carbon friendly.
“The Canada CIFAR AI Chair appointment got me more connected with the machine learning community.”
Gennady Pekhimenko, Canada CIFAR AI Chair, Vector Institute
Pekhimenko’s Toronto-based startup CentML, launched in 2022 with partners such as Radical Ventures, Amazon AWS and Microsoft Azure, provides system-level optimizations and tools for clients to improve the efficiency of their machine learning workloads. One such tool, Deep View, aids clients in choosing the most efficient hardware to run their AI/ML models, cutting costs and guesswork from the process.
Pekhimenko’s optimization research also targets carbon emissions reductions. CentML released a free prediction tool which helps their clients track and predict carbon load reductions. In addition, the company launched a service connecting AI/ML model developers with those who have compute to spare – think Facebook marketplace for AI/ML systems. CentML creates a micro-economy that scales up both the number of visible AI/ML models as well as companies’ ability to utilize their full computing power potential while cutting compute waste.
“The Canada CIFAR AI Chair appointment got me more connected with the machine learning community. It also provides extra stipends for students, which is important in a very competitive hiring market – especially against top U.S. schools,” says Pekhimenko, who has hired a number of former trainees for CentML.