Welcome!
Ph.D. Candidate in Computer Engineering | Reinforcement Learning and Bandit Algorithms
Welcome to my personal website where I plan on showcasing my work (as best I can). In the future I plan on using manim to create concise explainers for each of my work.
[Site still under construction but usable :) ]
About Me
I am a Ph.D. candidate at Texas A&M University, specializing in reinforcement learning. My academic journey includes a Master’s from Columbia University and a Bachelor’s from the National Institute of Engineering, Mysuru.
- Technical Skills: Python, PyTorch, TorchRL, Hugging Face, Weights & Biases, Slurm, Ray RLlib
- Coursework: Reinforcement Learning, Advanced Convex Optimization, Bandit Algorithms, Stochastic Systems, Analysis of Algorithms, Intro to Classical Analysis, Brain-Computer Interfaces
Research
Publications
- Risk-Averse Finetuning of Large Language Models (NeurIPS 2024)
- A Multi-Agent View of Wireless Video Streaming (INFOCOM 2024)
- EdgeRIC: Empowering Real-time Intelligent Optimization (NSDI 2024)
Patents
Ongoing Projects
Learning Algorithms for Restless Multi-Armed Bandits
Developing index policy learning algorithms for preference-feedback RMAB settings and providing regret bounds.
Autonomous Navigation of Comma Body Robot
Using RL and vision-based VQ-VAE architectures for autonomous navigation challenges.
Work Experience
Research Intern, AI Lab – InterDigital
Los Altos, CA | Mar 2021 - Aug 2021; May 2022 - Aug 2022
- Worked on deep learning-based video compression for live streaming applications.
- Improved existing models by adding convolutional-recurrent blocks and attention mechanisms into the inter-frame prediction branch.
- Built a reinforcement learning-based rate controller for deep learning-based video compression schemes.
Deep Learning Intern – Roche
Little Falls, NJ | May 2020 - Aug 2020
- Performed weakly supervised deep learning on gigapixel histopathology images for cancer detection using ResNet34 and RNN.
- Achieved an AUC metric of 99.3% on an in-house kidney cancer dataset.
- Presented results to the Data Science team at the Pharmaceutical Research and Early Development (pRED) Innovation Center.
Contact
Email: ujwald36@tamu.edu
LinkedIn: linkedin.com/in/ujwal-dinesha