Pushpendre Rastogi

pushpendre at gmail


I joined the Dialog State Tracking in Amazon Alexa in April 2019. I completed my Ph.D. in Computer Science at The Center For Language and Speech Processing, Johns Hopkins University. My advisor was Benjamin Van Durme. I TA'd graduate courses on representation learning and machine learning for three semesters during my Phd studies, and I received the George Sommerman Graduate Teaching Assistant Award with a cash award of $1000. I have reviewed for Transactions On Signal Processing-19, NEURIPS-19, ICML-19, ICLR-19, EMNLP-19, ACL-19, TPAMI-18, NeurIPS-18, KG4IR-18, EMNLP-18, ACL-18.

Selected Publications

See my google scholar profile for a complete list of publications.


Ph.D. and M.S. in Computer ScienceJohns Hopkins University2013-193.75/4.0
Thesis Topic: Representation Learning for Words and Entities. I presented new methods for unsupervised learning of word and entity embeddings from texts and knowledge bases.
Courses and Grades: Natural Language Processing (A), Machine Learning in Complex Domains (A), Stochastic Search & Optimization (B), Parallel Programming (A-), Principles of Programming Languages (A-), Combinatorial Optimization (A+), Introduction to Convexity (A-)
M.Tech. in Information and Communication TechnologyIIT Delhi2010-118.77/10
B.Tech. in Electrical Engg.IIT Delhi2006-108.86/10

Technical Notes

All notes are also hosted in a dropbox folder.
Note 1 Describes how the variance of an AB test can be reduced in the special case when we are comparing two policies with the same small-finite action space. [PDF]
Note 2 A video tutorial - primarily for my own understanding - about the difference between PnL and Cashflow, and how a company can have positive cash flow but still make loss, without raising debt. (you may need to download the video and play with VLC)
Note 3 [WIP] A visual proof of the UCB algorithm.


profile for Pushpendre on Stack Exchange, a network of free, community-driven Q&A sites