A Seattle-based machine learning startup, advised by a computer scientist who sold his company to Apple, just raised $3.9 million in seed funding. Now, investors hope the startup will follow in that company's footsteps.
OctoML on Wednesday announced the seed round led by Seattle-based venture firm Madrona Venture Group with participation from Silicon Valley-based Amplify Partners.
The startup is run by Luis Ceze, a longtime professor at the University of Washington's computer science school who consults for Microsoft. Ceze explains his company as being similar to iOS or Windows for machine learning, in the sense that it makes it easy for anybody to develop apps that take advantage of the technology.
"We're like an operating system for a machine learning model," he said.
Just as an operating system acts as a link between an application and hardware, OctoML acts as a link between machine learning models and devices, such as cars, phones and anything else with a processor.
It's a very technical concept, but it's the type of company that has lured success to Seattle-based companies before.
The Apple connection
Carlos Guestrin, now a senior director for AI and machine learning at Apple, is one of the creators of the company's flagship technology, and an adviser to the company. Guestrin sold his machine learning and artificial intelligence startup Turi to Apple in 2016 in a deal said to be worth some $200 million.
The acquisition helped establish Apple's presence in Seattle, which started with a 2014 acquisition of another Seattle startup and Madrona portfolio company Union Bay Networks.
Now, Apple is moving into a 630,000-square-foot office in Seattle's South Lake Union neighborhood, near Amazon and its Day One tower headquarters. Apple's website lists 84 job titles for which the company is hiring in Seattle, but it's not clear how many people the company plans to hire for each.
Will history repeat itself?
Madrona sees OctoML "as one in the vein" of Turi, spokeswoman Erika Shaffer said.
A machine learning model is basically a computer program trained to simplify a process — a foundational technology to what we think of as artificial intelligence It's one thing to build the model, Ceze said, but the model needs to be adapted so it can run in different ways, such as in data centers, cars, phone or health devices.
That's where OctoML's technology comes in.
OctoML's technology started as a research project out of the University of Washington's computer science school, led by a research team including Ceze and fellow professor Guestrin.
Ceze decided they should launch a company after 200 people showed up to a conference focused on the project last December. Ultimately, more than 270 people worked on the open-source project. Companies including Amazon, Microsoft and Facebook use the technology in their products.
What's it like to be in the room with a bunch of machine learning researchers?
"A lot of people speak really fast," Ceze said.
The company, based in downtown Seattle, has about 10 employees and has immediate plans following the funding round to hire another 10.