Featured Researcher - Deepak Agarwal
Researcher Deepak Agarwal was about to start a career in physics at the Presidency College in Calcutta when a professor ignited his interest in statistics. “He encouraged me to focus on statistics instead after he reviewed my answers to some probability theory questions in the physics entrance exam,” said Agarwal. And that’s exactly what he did.
Agarwal was born and raised in Calcutta, where he completed his undergrad and graduate studies, but moved to the United States in 1998 to pursue his PhD. After careful deliberation on school choices, he decided to attend the University of Connecticut because of the influence of Statistics Professor Alan Gelfand.
At the time, Agarwal admits he was an exception in his choice to continue his education. “Most of my friends were getting jobs because of the high tech boom,” said Agarwal. “A few of us decided to get our PhDs, realizing that we may be missing out on some lucrative opportunities in industry.” His thesis involved working on projects outside of the tech world. “I studied deforestation patterns in Madagascar,” said Agarwal. “I also worked with ecologists to understand Isopod settlement patterns in the Negev desert of Israel.”
When Agarwal completed his PhD in 2001, he accepted a position at AT&T Labs in New Jersey. He also received several other offers to stay in academia, but was influenced largely by the energy and opportunities at AT&T Labs. “There were large data problems and industry problems that appealed to me more,” said Agarwal.
It was unfortunate that AT&T’s financial fortunes started taking a turn for the worse soon after he joined AT&T Labs. “They laid off almost the entire machine learning team,” said Agarwal. Although things seemed to be crumbling around him, Agarwal worked on several high impact projects including telephony fraud detection and using telephone data to determine social connections between users. He also worked on developing methods for early detection of network problems.
After a “mass exodus” of employees at AT&T Labs in 2006, Agarwal left to take up his current position at Yahoo! Research. Most of his former colleagues went to Google – something he thought he would do as well. However, he felt that there was “more breadth and disciplines” at Yahoo! “Yahoo’s users and portals raise interesting statistical questions,” said Agarwal. ”There are more opportunities to do applied research and get involved in interesting problems relating to social networks [at Yahoo!].”
Without knowing anyone in the company, Agarwal joined Yahoo! Research in early 2006 and has never looked back. He is currently working on content optimization – figuring out through statistical algorithms what users want to see – to improve the click rates on Yahoo!. “This project has had an enormous impact on Yahoo!’s front page,” said Agarwal. “We’ve seen a surge in click rates and overall improvement in the quality of content.” He is also working on click-rate predictions for advertising – something he is modeling after the ecology work he did during his PhD. He enjoys the work mostly because of the scale and richness of data he works with – Internet user data as opposed to telephone user data at his previous position.
“When you wake up in the morning and look forward to going to work, I think it’s a good measure of how much you like your job,” said Agarwal.
When he’s not occupied with predicting click rates, Agarwal enjoys reading about Indian history, watching Bollywood and Hollywood movies and spending time with this 4-year old daughter, Naisha.
