Jul 2022 – Aug 2022 · 1 mo · Berkeley, California, United States

• Summer cluster on Deep Learning Theory

Jan 2022 – Nov 2023 · 1 yr 10 mos · San Diego, California, United States

• Deep Learning, Statistics, Optimization

Jun 2021 – Aug 2021 · 2 mos · New York, New York, United States

• Market Impact research for the Central Liquidity Book of the fundamental equities business

Apr 2021 – Jun 2021 · 2 mos

• Introduction to Machine Learning and Pattern Recognition (Statistics)

Jun 2020 – Sep 2020 · 3 mos · East Palo Alto, California, United States

• Worked on research problems in Natural Language Processing and Deep Learning

Mar 2020 – May 2020 · 2 mos · Palo Alto, California, United States

• Worked on NLP problems for the Matching Quality team at Amazon Search

Apr 2019 – Jun 2019 · 2 mos

• Principles of Feedback Control (Electrical and Computer Engineering)

Oct 2016 – Dec 2021 · 5 yrs 2 mos

• Theoretical and applied research in:
• Deep Learning
• Machine Learning
• High-dimensional Statistics
• Signal Processing
• Optimization
• Published articles in NeurIPS, ICML, AISTATS, JSAIT, ISIT, CCN, DeepMath

Mar 2016 – Sep 2016 · 6 mos · Los Angeles, California, United States

• Developed a framework for design and analysis of surge-pricing and capacity planning of electric vehicles charging infrastructures. Presented a paper at the American Controls Conference 2017 in Milwaukee.

Jul 2015 – Aug 2016 · 1 yr 1 mo · Mumbai Area, India

• I worked with the Systems and Controls group and was guided by Prof. Ankur A. Kulkarni on bridging the paradigms of Continuous optimization and Discrete Optimization. At its core, Continuous optimization problems are more amicable to solve computationally (perhaps using some gradient-based techniques), whereas solutions to Discrete optimization problems are more interpretable. We discovered some fundamental results bringing to light how each of these two, traditionally disparate paradigms, can benefit from the other.
• 1. Continuous Optimization benefiting from Discrete Optimization: (Published in Discrete Applied Mathematics journal)
• NP-hard combinatorial problems like the Independence Number problem can be posed exactly as Continuous Optimization problems using LCPs. This opens a new door for approximation algorithms. See the journal article for a more efficient SDP approximation.
• 2. Discrete Optimization benefiting from Continuous Optimization: (Published in Journal of Mathematical Economics)
• The phenomenon of free-riding (often described as the Tragedy of the Commons) is a classic problem in economics, and remains largely unresolved to this day. What makes the phenomenon persistent is the fact that a system with free-riding agents can be stuck in a Nash Equilibrium in several multi-agent games. In such scenarios, while the policy-makers may not be able to get rid of this phenomenon completely, we showed under several generic criteria these Nash Equilibria can be optimal. Our work has also opened several avenues to better understand the systemic flaws that lead to free-riding behavior among agents.

Jun 2012 – Apr 2015 · 2 yrs 10 mos · IIT Bombay, Mumbai

• Helped students develop problem-solving skills.
• Speech Processing (graduate-level course)
• Signals and Systems
• Partial Differential Equations
• Calculus (conducted extra discussion sessions in Hindi for freshmen facing a language barrier.)

May 2013 – Jul 2013 · 2 mos · Quebec, Quebec, Canada

• Learning hyperparameters to improve the performance of a 30 G-baud optical communication system. Performed simulations in MATLAB as well as worked on optical communication hardware.

Feb 2012 – Feb 2013 · 1 yr · Mumbai, Maharashtra, India