Portrait of Adam Kalai

Adam Kalai

Principal Researcher

About

I have been researching away at Microsoft Research New England (Cambridge, MA) since it’s founding in 2008 on various fun problems including 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 am co-chairing HCOMP 2017 (Conference on Human Computation and Crowdsourcing) October 24-26 in Québec City and NEML 2017 (New England Machine Learning Day) May 12 in Cambridge. Hope to see you there! Previously I co-chaired COLT 2010 (Conference on Learning Theory) and NEML 2012, 2013, 2015, and 2016.

Short Biography

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), Konstantina Christakopoulou (UMN), Abie Flaxman (CMU→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), 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)

 

Projects

Publications

Videos

Other

Kenneth C. Arnold, Kai-Wei Chang, and Adam Tauman Kalai

Improved Sentence Completions in Email Using Context

Abstract

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.

Draft