Pratik Chaudhari

I am an Assistant Professor in the Electrical and Systems Engineering department and a core faculty in the General Robotics, Automation, Sensing and Perception (GRASP) Laboratory. I hold a secondary appointment in the Computer and Information Sciences department and am a member of the Applied Mathematics (AMCS) graduate group, Penn Institute for Computational Science (PICS), Center for AI-enabled Systems: Safe, Explainable, and Trustworthy (ASSET) and Center For AI And Data Science For Integrated Diagnostics (AI2D).

Previously, I held a joint position as a Senior Applied Scientist at Amazon Web Services and a post-doctoral researcher at the California Institute of Technology in Computing and Mathematical Sciences. Currently, I am an Amazon Scholar at Amazon Web Services, where I focus on building large language models.

I defended my PhD thesis in the Computer Science department at University of California, Los Angeles in 2018 where I worked with Stefano Soatto in the UCLA Vision Lab. I have Engineer's (2014) and Master's (2012) degrees in Aeronautics & Astronautics from the Massachusetts Institute of Technology where I worked with Emilio Frazzoli at the Laboratory of Information and Decision Systems (LIDS). I was in the Aerospace Engineering department at IIT Bombay for my undergraduate studies until 2010.

I have worked extensively on self-driving cars in the areas of motion planning and computer vision at nuTonomy Inc (now Hyundai-Aptiv Motional).

Contact

pratikac at seas dot upenn dot edu

Address: Amy Gutman Hall 324G

Curriculum Vitae

Research Interests

My research group focuses on (a) theoretical problems in deep learning, and (b) visual perception in robotics. We conduct quite multi-disciplinary research along these lines. We use ideas from physics, information theory and neuroscience in addition to more classical perspectives from learning theory and computer vision. Everyone, from engineers and computer scientists to mathematicians and physicists will find something interesting to study in our group.

Research and Teaching Statements A synopsis of my recent research and teaching efforts. You can also read a slightly dated research statement from 2017 to see how these ideas have matured. An aspirational blogpost that stems from some recent work is here. A blogpost by the School of Engineering on a recent result is here. Another blogpost on broad research directions and teaching is here.

NeurIPS 2020 Workshop on Deep Learning through Information Geometry We organized a workshop to brainstorm how ideas in information geometry and information theory inform questions of optimization and generalization in deep learning. You can find the recorded talks and accepted submissions at here.

Winter School on “Quantitative Systems Biology” at the International Center for Theoretical Physics (ICTP) in Trieste, Italy You can find the recorded talks here

New Frontiers in Learning, Control, and Dynamical Systems” workshop at ICML 2023 Link

Highlights

Brief overview

A picture of the prediction space of deep networks

Does the data induce capacity control in deep learning?

Unraveling the mysteries of stochastic gradient descent for deep neural networks

Learning with few labeled data

Research Group

Postdoctoral Scholars

Doctoral students

Doctoral Alumni

Master's thesis students Erik Jagnandan (Robo 2025), Zach Osman (Robo 2025)

Undergraduate students Siming He (CS/MNT, 2021-23) Alexander Kyimpopkin (ESE/Robo 2024-25)

Alumni

Master's thesis students Daiwei Chen (ESE. Now Wisconsin EE PhD), Wei-Kai Chang (Scientific Computing. Now Purdue CS PhD), Megharjun Nanda (Robo. Now co-founder at Sahay AI), Yingtian Tang (CIS. Now EPFL Neuroscience PhD), Haoran Tang (CIS. Now Purdue CS PhD), Ashish Mehta (Robo MS, now at Qualcomm), Christopher Hsu (Robo MS, now at ARL/Penn EE PhD), Wenbo Zhang (Robotics MS, now at grayscale.ai), Sebastian Peralta (Physics/ESE BS, Robo MS, now at Amazon)

Undergraduate students Brian Liu (2022), Aalok Patwa (2022), Max Wang (2022), Sherry Chen (EE/MNT BS, CIS MS, now at Chef Robotics), William Qian (CIS BS, Physics MS, now at Harvard Physics PhD), Andrew Zhu (CS, 2021-23) Emily Paul (CS, 2021) Suhaila Shankar (CS/Physics, 2022-23), Yitzy Tanner (BE/Physics, 2022-23), Alicia Sun (2023), Alan Zhu (2023), Parsa Idehpour (2024), Ming Qi (2024), Yazlin Moujalled (2024), Adithya Selvakumar (2024)

Teaching

Fall: ESE 546 Principles of Deep Learning
Syllabus
Notes: Fall 2024
Accessible notes: Fall 2023

Spring: ESE 650 Learning in Robotics
Syllabus
Notes: Spring 2024

Publications

See Google Scholar for the latest list of publications and citations.

Theses

  1. The Hidden Geometry of Learning, Jialin Mao, University of Pennsylvania, 2025 (PDF)
  2. The Principles of Learning on Multiple Tasks, Rahul Ramesh, University of Pennsylvania, 2025 (PDF)
  3. Fair and Generalizable Machine Learning for Neuroimaging, Rongguang Wang, University of Pennsylvania, 2024 (PDF)
  4. An Information Geometric Picture of the Space of Tasks, Yansong Gao, University of Pennsylvania, 2023 (PDF)
  5. A picture of the energy landscape of deep neural networks PhD, Computer Science, University of California, Los Angeles, 2018 (PDF)
  6. Algorithms for autonomous urban navigation with formal specifications Engineer, Aeronautics-Astronautics, Massachusetts Institute of Technology, 2014 (PDF)
  7. Incremental sampling based algorithms for state estimation SM, Aeronautics-Astronautics, Massachusetts Institute of Technology, 2012 (PDF)