Shreyas Chaudhari

Shreyas Chaudhari

PhD Student in Computer Science

University of Massachusetts Amherst

About Me

I am a first-year PhD student in Computer Science at UMass Amherst advised by Bruno Castro da Silva and Phil Thomas at the Autonomous Learning Lab. I work on reinforcement learning, often using tools from deep learning. I have also worked on various related areas like uncertainty quantification, natural language processing and multi-armed bandits, details about which can be found in the Projects section.

Education

  • PhD in Computer Science, Ongoing

    University of Massachusetts Amherst

  • MS in Electrical and Computer Engineering, 2020

    Carnegie Mellon University

  • B.Tech. in Electrical Engineering, Minor: Machine Learning, 2019

    Indian Institute of Technology Madras

Professional Experience

 
 
 
 
 

Software Engineering Intern

Uber ATG, Perception/Prediction Team

Jun 2020 – Aug 2020 Pittsburgh
  • Derived loss functions for confidence-calibrated neural network predictions for image classification and object detection
  • Shifted from two-step post-processing calibration methods, benchmarking one-stage object detectors (FAIR’s RetinaNet)
  • Achieved improved calibration scores on image classification along with occasional improvement in model performance [informal report]
 
 
 
 
 

Securities Strategist Intern

Goldman Sachs, Franchise Analytics Strategy & Technology Team

May 2018 – Jul 2018 Bengaluru, India
  • Analyzed the trade data for European markets post MiFID II regulation, gauging the monetary value of the dataset
  • Conducted trade volume and market share driven analysis indicating ∼40% increase in market coverage by the data
  • Modeled quoting patterns with GMMs using pre-trade data feed to draw correlations across aggressiveness of quoting
 
 
 
 
 

Machine Learning Intern

DeTect Technologies, Machine Learning Team

May 2016 – Jul 2016 Bengaluru, India
  • Developed code for the flagship product GUMPS which detects defects and their growth rate in oil refinery pipelines
  • Implemented idea analyses reflected waveforms of ultrasound waves for defect detection

Projects

*
All Machine/Deep Learning Reinforcement Learning Multi-Armed Bandits Natural Language Processing Computer Vision Data Science Course Projects

Training objective functions for confidence-calibrated predictions from neural networks

Internship Project, Uber ATG
Shreyas Chaudhari
PDF

Off-Dynamics Reinforcement Learning

Research project, CMU
Ben Eysenbach*, Shreyas Chaudhari*, Swapnil Asawa*, Ruslan Salakhutdinov, Sergey Levine
PDF Code Slides

Off Policy Meta-Reinforcement Learning

Course Project: Probabilistic Graphical Models, CMU
Karmesh Yadav, Shreyas Chaudhari, Vivek Pandit, Shangbang Long
PDF Code

Classification of gene expression in scRNA-seq data

Course Project: Machine Learning, CMU
Shreyas Chaudhari, Vishwak Srinivasan, Shantanu Gupta
PDF Slides

Multi-Armed Bandits with Correlated Arms

OPAL lab, Carnegie Mellon University
Samarth Gupta, Shreyas Chaudhari, Gauri Joshi, Osman Yağan
PDF Code

Classical Bandit Algorithms to the Structured Bandit Setting

OPAL lab, Carnegie Mellon University
Samarth Gupta, Shreyas Chaudhari, Subhojyoti Mukherjee, Gauri Joshi and Osman Yağan
PDF Code

Second-Order Methods for Policy Search in Reinforcement Learning

B.Tech. thesis, IIT Madras
Shreyas Chaudhari
PDF

Risk-Sensitivity in Multi-Armed Bandits

Course Project: Multi-Armed Bandits, IIT Madras
Shreyas Chaudhari, Monisha Jegadeesan
PDF Code

Multi-Hop Question Answering using GCNs

Research project, IIT Madras
Shreyas Chaudhari, Sahana Ramnath
PDF

Risk-Sensitive Reinforcement Learning: A Comparative Analysis

Course Project: Reinforcement Learning, IIT Madras
Shreyas Chaudhari, Monisha Jegadeesan
PDF Code

Equivalence of GANs and Zero Sum Games

Course Project: Game Theory
Shreyas Chaudhari, Abhiman N.
Slides

Defaulting Customer Prediction

Big Data Challenge, American Express
Shreyas Chaudhari, Ganga Meghanath
Slides

English Text to Vernacular Speech

AWS Deep Learning Hackathon
Shreyas Chaudhari, Monisha Jegadeesan

Energy-Delay-Distortion Problem

Tata Institute of Fundamental Research, Mumbai
Rahul Vaze, Shreyas Chaudhari, Akshat Choube, Nitin Aggarwal
PDF

Payload Team

IITMSAT: Student satellite project, IIT Madras

Publications

Off-Dynamics Reinforcement Learning: Training for Transfer with Domain Classifiers

ICLR 2021
Ben Eysenbach*, Shreyas Chaudhari*, Swapnil Asawa*, Ruslan Salakhutdinov, Sergey Levine
PDF Code

Multi-Armed Bandits with Correlated Arms

Theoretical Foundations of Reinforcement Learning workshop, ICML 2020
Samarth Gupta, Shreyas Chaudhari, Gauri Joshi, Osman Yağan
PDF Code

A Unified Approach to translate Classical Bandit Algorithms to the Structured Bandit Setting

IEEE Journal on Selected Areas in Information Theory | ICASSP 2021
Samarth Gupta, Shreyas Chaudhari, Subhojyoti Mukherjee, Gauri Joshi and Osman Yağan
PDF Code

Energy-Delay-Distortion Problem

Oral presentation, National Conference of Communications 2018
Rahul Vaze, Shreyas Chaudhari, Akshat Choube, Nitin Aggarwal
PDF

Contact