Dr. Rohit Khare, 2003 ISR Graduate

“Has it really been fifteen years?!,” Rohit Khare exclaimed.  “I first heard about Professor Richard Taylor’s unique research team at UC Irvine in the mid-90’s, from his students working at the World Wide Web Consortium — and that pretty much determined the course of my career ever since.”

Indeed, within a few years the future Dr. Khare had moved from MIT to Irvine to enroll as a new graduate student, collaborated on several new research themes at ISR, and helped bring them to market as an entrepreneur.  His first venture-backed startup, KnowNow, was a direct spinoff of his doctoral research; while his latest startup, Ångströ, was recently acquired by Google.  In between, he helped re-launch CommerceNet as a research lab and incubator around the theme of decentralization, something he began investigating seriously in Prof. Taylor’s research group and continues to drive his work today.

Centralized systems are easy to identify, since there’s one person, one company, or one computer ‘in charge.’  Distributed systems make that model practical by sharing control across a network of machines, nowadays made popular as ‘cloud’ computing.  Decentralized systems, by contrast, promise ‘software that works the way society works:’ acknowledging that in a peer-to-peer system, the best we can do is track what everyone else is doing and stop waiting for the “true” value for prices, inventories, contact information, and the like.

“Consider your humble ‘little black book,’” Dr. Khare offered by way of example.  “You could jot it all down in a central Rolodex, but you probably distribute copies across your PC, your laptop, and your phone and then try to keep them all in sync.”  “In the real world, though, that contact info you stashed away may already be out of date.  The really interesting data is your ‘social graph,’ and that lives in pieces on Facebook, LinkedIn, Google+ and lot of other sites where each of your hundreds of contacts are constantly updating their status,” he explained.

His solution to that problem is what caught Google’s eye as they were assembling the team behind Google+.  Part of the reason they acquired Ångströ last Fall was its software for coping with the fact that people may have many relationships with overlapping communities on multiple social networks.  “Rather than building yet another address book to trap your data, we decided to put users back in control by knitting together your existing social graph, on-the-fly,” he added.

In the mid-90’s, while working on Internet standards at W3C with Tim Berners-Lee and later at MCI with Vint Cerf, he kept running into some “deeply informed and deeply committed” graduate students from Prof. Taylor’s group, including Dr. Roy T. Fielding, of HTTP, Apache, and REST fame and ISR Prof. E. James Whitehead, who helped extend the Web to support Distributed Authoring and Versioning (WebDAV). “I was impressed that there was a research group out there that took standards and protocol design at the application layer seriously,” he recalled. So he began his studies in earnest by enrolling at the Bren School of Information and Computer Sciences at UC Irvine in 1997.  As the past editor of the World Wide Web Journal, Khare had a broad knowledge of issues facing Web developers from fonts to payment protocols, so as Prof. Taylor recalls “It is hard to overstate what a challenge it was to keep Rohit focused!”

By 1998, he worked to help convene the Workshop on Internet-Scale Event Notification (WISEN) at UCI.  He presented a survey of the field that covered dozens of new and established real-time update protocols in collaboration with Adam Rifkin, a colleague from his undergraduate days at Caltech.

Part of earning a graduate degree is learning how the research funding process works.  “After the success of WISEN’98, Prof. Taylor started suggesting new themes our team could investigate with respect to wireless networking, particularly real-world applications using “low-tech” pager networks,” Dr. Khare recalled.  So when the NSF rejected their next research proposal as “too commercial,” it was no surprise that in the full-flower of the dot-com boom, Dr. Khare took that as a sign to ‘drop out’ and launch KnowNow with his friends, fellow UC Irvine alumnus Dr. Peyman Oreizy and Mr. Rifkin.

According to Mr. Rifkin, “Ultimately, we raised over $50M over 8 years pursing the Enterprise market for real-time event notification across the Internet, when it turned out that the ‘killer app’ for advanced Web apps turned out to be consumer applications such as Gmail and Twitter.  Today, the sort of capabilities we struggled to emulate by writing a miniature Web server in JavaScript are now baked right into HTML5 and the BrowserChannel — there’s a big lesson to be learned about the difference between being right, and being at the right time and the right place.”

For Dr. Khare, though, the KnowNow journey was shorter than he expected — after another year as CTO, he decided to resign from the Board and return to pursuing his doctorate at Irvine full-time.  Over the following two years, he finished his dissertation, building upon Dr. Roy Fielding’s seminal work on the REpresentational State Transfer (REST) architectural style to describe how technology like KnowNow’s could help developers of decentralized systems, winning a Best Paper award at the 2004 International Conference on Software Engineering (ICSE).

By 2006, Dr. Khare had spun off his own startup to revisit the market for publish/subscribe event notification in the age of Facebook.  Ångströ asked, “What if people could follow news about people and companies that matter to them?”  Obviously, existing approaches such as Google Alerts are flummoxed by the existence of people with the same name, but with the advent of LinkedIn profiles and the like, Dr. Khare began developing algorithms to disambiguate names based on user feedback — and to rank which topics were most interesting, too.

As the Ångströ team iterated through different products and services for social information mining, ranging from photos to games to marketing analytics, one common platform component that kept improving was our decentralized social graph matching engine.  Rather than building its own universal list of all the newsworthy people in the world, Ångströ imported bits of users’ social graphs from other sites (as well as the occurrence of news articles tying people together in the same context) to build a unified model, able to map a number to a name to a LinkedIn title, Facebook status, Flickr photo, and more.

Today, at Google+, Dr. Khare is responsible for similar aspects of their friend finding and suggestion tools.  He is committed to helping to build a more open social Web by contributing to a Google+ Platform and APIs that help other developers accommodate a multilateral marketplace, not just tying their identity and event systems to one company or another.  “At Google+ we talk about fixing ‘sharing on the web’ so it works ‘like it does in the real world’ — pretty much a straight line from the mission Dr. Taylor and I worked out years earlier, to ‘build software that works the way society works.’  And I can see myself still working to the same goals for years to come.”