The Seattle-based VC firm Madrona Venture Group has led a seed investment in a University of Washington startup called PetraVM. The deal is worth $1.5 million, and includes investment from Seattle-based WRF Capital. Founded by professors Mark Oskin and Luis Ceze of the UW’s computer science and engineering department, PetraVM is out to improve the cost and reliability of software that people use every day—software that increasingly runs on parallel-processing hardware.
Last week, I sat down with Matt McIlwain, managing director at Madrona, to hear how the deal came about. Madrona has a strong track record of investing in UW spinoffs—a dozen or so, including Impinj, Skytap, Physware, and Farecast. McIlwain himself is on the board of many local startups including Skytap, Apptio, Extrahop, Smartsheet, and Isilon Systems.
The latest funding has its roots in Madrona “office hours.” This is an outreach program whereby Madrona VCs spend a few mornings per year in the UW computer science building, meeting with faculty and students from all engineering departments. McIlwain says they talk about the process of starting companies, how venture capital and angel investments work, and the like.
In January 2008, McIlwain sat down with Ceze and Oskin at Madrona office hours, and they hit it off, talking about approaches to improving software development. “Hank Levy [chair of computer science and engineering] had mentioned them to me,” says McIlwain. “He said, ‘It’s great that they’re on your schedule.'”
Over the next six months, McIlwain says, he did a “deeper dive” into whether the founders’ ideas could form the basis of a viable company. Oskin and Ceze are both experts in computer architecture. Oskin has led research in on-chip networks and quantum-scale architectures, while Ceze’s group has recently developed an approach called “deterministic shared memory multiprocessing.”
Despite these technical terms, the purpose of Oskin and Ceze’s startup is quite simple. Its goal is to “make it cheaper to make reliable, multi-threaded software,” says Ceze. Most software that people use these days is multi-threaded—meaning it runs on multiple processors that need to interact efficiently—and this “fundamentally changes the way you write code,” Ceze says. “That is much harder than software we’ve been writing for decades.”
Typical examples of multi-threaded software include Firefox, Microsoft Office, and database applications. So of course, PetraVM faces plenty of competition from companies such as Cilk Arts, an MIT spinoff in the Boston area that my colleague Wade profiled in late 2007.
So who else is watching this space locally? “Oh, maybe a little company called…Microsoft?” McIlwain offers. Indeed, it seems if PetraVM is successful in lowering software costs, then giants like Microsoft, Google, and Amazon might all beat a path to its door. “These companies are riding multi-threaded code for some time now,” says Ceze.
Oskin is now full-time at PetraVM, while Ceze is a consultant. The startup has hired a team of three engineers, with Madrona’s help. The new funds are being used to expand the team and ramp up product development. “We expect this money to last a little over a year,” says Ceze. “We’re hiring more now. There’s lots of engineering going on.”
