I have been researching away at Microsoft Research New England (Cambridge, MA) since its founding in 2008 on various fun problems including machine learning, program synthesis (getting computers to write programs), crowdsourcing, accessibility, and computational humor, to name a few. Here’s an NPR story about our recent work on getting computers to be less biased (or try the demo we made recently at a hackathon).
Previously, I served as an Assistant Professor in Computer Science at Georgia Tech and TTI-Chicago. I was extremely fortunate to receive a PhD at CMU from the ingenious Avrim Blum (where I was also supervised by Randy Pausch for a time), followed by an NSF postdoc at MIT under the wise guidance of Santosh Vempala.
I’ve chaired and co-chaired a number of conferences and meetings including NEML (New England Machine Learning Day), HCOMP 2017 (Conference on Human Computation and Crowdsourcing), and COLT 2010 (Conference on Learning Theory).
I’m a big fan of the safe password strategies of Blum, Samadi, and Vempala.
Adam Tauman Kalai received his BA (1996) from Harvard, and MA (1998) and PhD (2001) under the supervision of Avrim Blum from CMU. After an NSF postdoctoral fellowship at M.I.T. with Santosh Vempala, he served as an Assistant Professor at the Toyota Technological Institute at Chicago and then at Georgia Tech. He is now a Principal Researcher at Microsoft Research New England. His honors include an NSF CAREER award and an Alfred P. Sloan fellowship. His research focuses on machine learning, human computation, and algorithms.
Students & Teaching
Previous students/interns: Jake Abernethy (Berkeley→Georgia Tech), Ken Arnold (Harvard), Pranjal Awasthi (CMU→Rutgers), Nina Balcan (CMU→CMU), Harry Bovik (CMU→CMU), Danielle Bragg (UW), Elica Celis (UW->EPFL), Konstantina Christakopoulou (UMN), Abie Flaxman (CMU→UW), Brittany Fiore-Gartland (UW->UW), Varun Kanade (Georgia Tech→Oxford), Katrina Ligett (CMU→Caltech), Azarakhsh Malekian (UMD→Toronto), Brendan McMahan (CMU→Google), Aditya Menon (UCSD→NICTA), Ankur Moitra (MIT→MIT), Claire Monteleoni (MIT→GWU), Peter Organisciak (UIUC→HathiTrust), Aaron Roth (CMU→UPenn), Omer Tamuz (Weizmann→Caltech), Jason Tsay (CMU), Shubham Tulsiani (IIT→Berkeley), Duru Turkoglu (U. Chicago→DePaul), Shyam Upadhyay (UIUC), Greg Valiant (Berkeley→Stanford), Elad Verbin (Tel Aviv University→Aarhus), Miaomiao Wen (CMU→Coursera), and Kuat Yessenov (MIT→Google).
Spring 2008: Game Theory and Computer Science, (Georgia Tech)
Fall 2006: Machine Learning Theory (Weizmann Institute)
Autumn 2004: Online Algorithms (University of Chicago)
Improved Sentence Completions in Email Using Context
Many people find text autocompletion helpful, but current systems are limited to predicting only one or two words at a time. We consider the “Megacomplete” text prediction problem of suggesting substantial chunks of text (e.g., sentences or longer), in situations where the context makes a certain completion highly likely. We design a system that learns to identify good completions based on an individual’s corpus of thousands of sent email messages, based on features of the context (e.g., subject) and the text typed so far. We evaluate the system based both on exact text matches and on human evaluation of completion acceptability. A companion paper studies usability factors of the system.