Education

Carnegie Mellon University

Relevant Graduate Coursework (4.0 GPA, all A grade):

• Modern Convex Optimization (Tepper Operations Research PhD Class)
• Advanced Statistical Learning Theory (In progress)
• Theory of Markov Processes (In progress)

The University of Texas at Austin

Relevant Graduate Coursework (4.0 GPA, all A grades):

• Statistical Machine Learning
• Stochastic Analysis
• Probability and Stochastic Processes I
• Learning-Based Optimal Control
• Reinforcement Learning: Theory/Practice
• Generative Models in Machine Learning
• Applied Machine Learning
• Machine Learning on Real-World Networks
• Human Signals: Sensing and Analytics

Research Experience

Recent Projects

Convex Social Network Design

• First work to build real-world networks emphasizing all three properties: high clustering and locality (small-world), degree heterogeneity and hub presence (scale-free), and global connectivity.
• Formulated the problem as a convex program solvable both centrally and via a decentralized edge potential game, with the potential equal to the Lagrangian of the convex program.
• Proved that the Nash equilibrium of the game is the global optimizer of the program (converging to the KKT point).
• Designed a projected primal–dual subgradient method to solve the resulting saddle-point optimization of the Lagrangian and proved convergence.

Combinatorial Bandits for Vaccine Uptake Prediction

• Provided provable regret guarantees and empirical support in both the semi-bandit and full-bandit feedback cases.

Restless Boosting Bandits for Maternal Care

• With each arm representing an independent MDP, solved the problem of optimizing maternal care service calls using an adaptive boosting bandit algorithm.

Solving a Disinformation Game via Stackelberg Policy Gradient

• Developing strategic policies in adversarial environments to counter disinformation by modeling the interaction between defenders and adversaries as a Stackelberg game.

Data Curation for Speech Disfluencies via Generative Models

• Curated and analyzed datasets focused on disfluencies in spoken language, using generative approaches to enrich training corpora.

Monolingual vs. Bilingual Speakers Analysis

• Compared speech patterns via logistic regression (controlling for linguistic structure) to highlight key differences in fluency, disfluency markers, and language-switching behavior.