A Carnegie Mellon University spinout that wants to make it easier for businesses to use machine learning software has obtained $15 million in initial venture capital funding.
Petuum Inc. will leave the university and set up shop in Pittsburgh while enlarging its staff of 15 people to 30 to 40 by next spring, said founder and CEO Eric Xing, who is also a professor of computer science at CMU.
The company’s first products are expected for release early next year.
“It’s really for the goal of making [machine learning] a commodity, rather than keeping artificial intelligence as a black box or trade secret,” Mr. Xing said. “You don’t have to build your own.”
A kind of artificial intelligence that allows software to learn information without explicit programming, machine learning is only used by about 10 percent of big businesses, Mr. Xing said.
Machine learning software, which is key to such innovations as self-driving cars, speech recognition and natural language processing, is mostly handcrafted by engineers for specific applications, Mr. Xing said.
The software platform being developed by Petuum will “democratize” machine learning, making it more widely available with tools that are easier to access.
The technology could open up new possibilities in banking, finance, manufacturing, energy and other industries.
“There’s a huge need to make better sense out of that data,” Mr. Xing said. “We’d really love to grow this project in Pittsburgh.”
Google and Amazon each use advanced machine learning but not in the way planned by Petuum.
Among the company’s early products will be a way to assess disease risk and predict hospital readmission rates by analyzing electronic medical records and searching for patients with similar conditions.
Until recently, Mr. Xing directed the Center for Machine Learning and Health at CMU, which is part of the Pittsburgh Health Data Alliance, a collaboration of CMU, University of Pittsburgh and UPMC.
Kris B. Mamula: kmamula@post-gazette.com or 412-263-1699.
First Published: November 29, 2016, 5:00 a.m.
