Featured Researcher - Kishore Papineni
Kishore Papineni grew up in a “small” town in India. With a population of over a half a million people, it seemed like his childhood in Guntur positioned him for a career in engineering. “That’s what the top of the class did – either engineering or medicine,” he says.
His interest was in control theory – a branch of engineering and mathematics that uses feedback to control the desired effect on the output of dynamical systems. He laughs as he reminisces. “I had a bad teacher for this class, and despite that I found the subject fascinating,” he says. That’s when he realized he was serious in pursuing the field.
Papineni specialized in feedback control theory for his PhD at Rice University in Houston. He had aspirations of teaching at first, but by the end of his thesis, he was looking to work for an industry research lab instead. He describes himself as a “pen and paper theorist” because of his focus on theory versus practical applications. In his thesis, he disproved some well-known theorems in his field that have been taught for decades. He says working on the foundational aspects of systems theory was not the mainstream for PhDs in his field. “In my entire graduate career, I probably wrote a total of 200 lines of C-code,” he says. “I never thought I would get into practical work because I was always a theoretician.”
After he completed his PhD, Papineni joined IBM T.J. Watson Research Center, where he was a part of the speech recognition group. “It was a total shift from what I did for my PhD,” he says. The unexpected choice in profession led to a broad range of roles, including work in natural language understanding, dialogue management, and statistical machine translation – a role that he found especially challenging and fulfilling. He led the effort in primarily translating foreign language documents into English. “It provides a rich playground to apply all sorts of natural language processing technologies and machine learning techniques,” he says. He describes his 11 years of service at IBM as “happy” and “fulfilling.”
Papineni left IBM and joined Yahoo! Research in July 2006. He admits to being nervous at first about leaving IBM to come to Yahoo!. So why did he leave IBM if he was so happy there? “It’s really about impact,” he says, “along many dimensions: myself, my employer, my professional community, and on the world at large.” He explains that Yahoo! provides great potential in each of these dimensions with a chance to “touch the online experience of millions of users all over the world.” His roles at Yahoo! include modeling placement of pay-per-click advertisements, modeling guaranteed delivery of banner advertisements, and running the machine learning group. The machine learning group works on an array of advertisement and audience related projects and his goal is to achieve the right balance between theoretical and practical work for the group.
In his spare time, Papineni enjoys browsing and acquiring plants, trees, and shrubs with his wife. “We like pretty much anything that flowers,” he says. Landscaping their yard has been a favorite pastime.
He lives in Carmel, a picturesque town about 50 miles north of New York City and faces a long commute each way to work. But with wireless connectivity on the train, his effective commute is less than 40 minutes. Despite the long commute, he couldn’t be happier. He has “absolutely no regrets,” he says. “I’m having a ball.”
