Alok Shah

I am a visiting researcher in the Machine Learning Department at Carnegie Mellon University, hosted by Max Simchowitz. This fall, I will begin my PhD at MIT IDSS.

I previously studied Electrical Engineering and Computer Science at the University of Pennsylvania, where I was gratefully advised by Nikolai Matni in the GRASP Laboratory. Outside of class, I led MLR@Penn, Penn's undergraduate AI and machine learning research organization and community, and served on the Student Advisory Board for the Wharton AI & Analytics Initiative.

In my free time, I enjoy learning about human geography, trying new foods, and burning off those calories by playing pickup soccer.

Email  /  CV  /  Scholar  /  Twitter  /  Github

profile photo

Research

I am broadly interested in the theory and practice of deep learning. Recently, I have been thinking about how to adapt our favorite training recipes to handle long-horizon learning problems, which frequently arise in language modeling and robotics.

Projects
Horizon Dynamics
Khush Gupta*, Faraz Rahman*, Aryan Roy*, Alok Shah*, Ethan Yu*
Senior Design Project
Most Innovative and Techincal at Jerome Fisher Management and Technology Summit; CIS Senior Design Winner
Developed an interpretability and observability platform to autonomously debug robot learning systems
An Information-Theoretic Analysis of Continual Learning with Structured Memory
Dominic Olaguera-Delogu*, Nikhil Kumar*, Alok Shah*
ESE 6740 Final Project
Using Grassmanian subspace packing, we prove that external memory grows sublinearly in the number of tasks when continually learning from a rank-r linear Gaussian family.
Dropout, Johnson-Lindenstrauss, Low-Rank Bias, and Generalization
Alok Shah*, Mohul Aggarwal*, Khush Gupta*
CIS 6770 Final Project
An alternate perspective on the generalizing behavior of dropout through the lens of random projections and low-rank bias
Provable Multitask Learning with Task-Adaptive Priors
Alok Shah*, Sidhant Srivastava*, Tara Kapoor*
CIS 7000 Final Project
By extracting task-dependent statistics, we learn a prior that provably nudges the model toward representations that generalize better to new tasks.
Conformal Actor-Critic: Distribution-Free Uncertainty Quantification for Offline RL
Alok Shah*, Nikhil Kumar*, Khush Gupta*, Mohul Aggarwal*
CIS 6200 Final Project
Integrated conformal prediction into offline reinforcement learning, providing statistically robust uncertainty quantification to curb overestimation bias of Q-values
Deep Compression with Adversarial Robustness via Decision Boundary Smoothing
Alok Shah, Michael Shao
ESE 5390 Final Project
Smoothed adversarial retraining during compression produces compact models with stronger robustness
AlfLLM: Limitations on LLMs as Reward Function Surrogates
Alexander Kyimpopkin, Alok Shah, Dominic Olaguera-Delogu
ESE 6500 Final Project
Investigated the efficacy of using large language models as surrogate reward functions for reinforcement learning in the ALFWorld environment.

Selected Coursework

* denotes graduate level; ** denotes doctoral level

  • Honors Analysis in Several Variables
  • Discrete Mathematics
  • Linguistics
  • Data Structures and Algorithms
  • Advanced Linear Algebra*
  • Probability*
  • Computer Systems
  • Machine Learning*
  • Real Analysis
  • Topology*
  • Differential Geometry*
  • Operating System and Design
  • Ethical Algorithm Design*
  • Convex Optimization**
  • State Estimation, Control, and Reinforcement Learning**
  • Deep Learning*
  • Automata, Computability and Complexity
  • Analysis of Algorithms*
  • Uncertainty Quantification**
  • Learning for Dynamics and Control**
  • Hardware/Software Co-Design for Machine Learning*
  • Stability in Optimization and Statistics**
  • Bayesian Optimization**
  • Randomized Algorithms and Numerical Linear Algebra**
  • Numerical Optimization for Data Science and Machine Learning
  • Statistical Topics in Large Language Models**
  • Statistical Mechanics**
  • General-Purpose GPU Programming Architecture**
  • Physical Intelligence**
  • Human-Computer Interaction*
  • Acceleration and Hedging in Optimization**
  • General Relativity*
  • Information Theory**
  • Sampling and Optimal Transport**
  • Senior Design

Teaching

I am currently faculty at the Pennsylvania Governor's School for the Sciences at Carnegie Mellon University where I teach Learning for Robotics.

I enjoy teaching and have served on staff for several courses.

* denotes graduate level; ** denotes doctoral level; ^ denotes head teaching assistant

  • Mathematics of Machine Learning (Eric Wong)
  • Machine Learning*^ (Surbhi Goel, Eric Wong, Jake Gardner, Lyle Ungar)
  • Statistics for Data Science** (Hamed Hassani)
  • Deep Learning* (Pratik Chaudhari)
  • State Estimation, Control, and Reinforcement Learning** (Pratik Chaudhari)
  • Convex Optimization** (Nikolai Matni)

Shamelessly forked from Jon Barron.