Hi!
I am a PhD student at the Machine Learning Department, Carnegie Mellon University.
My interests lie in the development of robust, generalizable model-based RL algorithms for real-world control, using limited data.
Previously, I worked at Blue River Technology as an ML engineer after graduating from CMU with a Masters in ML. During my Masters, I pursued research in model-based reinforcement learning with Prof. Fei Fang, and with Prof. Jeff Schneider at the Auton Lab at CMU.
Prior to starting my Masters, I was a Post-Baccalaureate Research Fellow at RBCDSAI, IIT Madras, advised by Prof. Balaraman Ravindran. My research at RBCDSAI spanned model-based RL and robust game-theoretic RL with applications in real-world problems, which included combating animal poaching.
I did my undergraduate thesis at the Robotics Insitute, CMU, advised by Prof. Cameron Riviere at the Surgical Mechatronics Lab, on real-time blood vessel avoidance for neuro-surgical procedures using Micron, a surgical robot. I graduated with a B.E. (Hons.) in Computer Science and an M.Sc. (Hons.) in Economics from BITS Pilani, Goa, India, in 2020.
I also interned in the autonomous driving industry, at Blue River Technology in summer 2023 working on 3D bird's eye view object detection for collission avoidance in autonomous tractors.
For more information, please check out my CV/contact me here!
Experience
-- Developed a semi-supervised learning algorithm to improve semantic segmentation model performance in low-labelled-data regimes.
-- Developed and experimented with active-learning strategies to down-sample large datasets for semantic segmentation on agricultural images.
-- Developed a hierarchical latent variable model, Multi-Agent Bi-Level Model (MABL) for multi-agent RL (paper).
-- Extended MABL to learn temporally abstracted action and state sequence representations for RL.
-- Developing Denoising Probabilistic Diffusion Models to learn stochastic dynamics from real-world, offline RL data.
-- Designed a real-time 3D Transformer framework for object detection on autonomous tractors in the field.
-- Constructed large-scale pseudo-label datasets and pipelines applying classical computer vision.
-- Developed, CombSGPO, which uses game theory and RL to combat wildlife poaching. (paper).
-- Developed an RL approach for empirical comparison with evolutionary algorithms in Green Security Games (GSGs), collaborating with Prof. Jacek Mańdziuk. (paper).
-- Developed an optimization framework to achieve robust equilibrium performance in Markov games.
-- Developed a real-time virtual fixture strategy for Micron, a handheld surgical robotic tool, to avoid blood vessels during neurosurgery and conducted real-world trials. (thesis).
-- Worked towards vessel avoidance using Micron by implementing accurate and fast deep-learning algorithms for real-time, intra-operative blood vessel segmentation.
Publications
Bi-level Latent Variable Model for Sample-Efficient Multi-Agent Reinforcement Learning
, Stephanie Milani, Fei Fang, Balaraman Ravindran.
pdf (Under Review)
LaVa: Latent Variable Models for Sample Efficient Multi-Agent Reinforcement Learning
, Elizabeth Bondi, Fei Fang, Balaraman Ravindran.
Reinforcement Learning and Decision Making (RLDM) 2022
pdf
Evolutionary Approach to Security Games with Signaling
Adam Zychowski, Jacek Ma ́ndziuk, Elizabeth Bondi, , Milind Tambe and Balaraman Ravindran.
International Joint Conferences on Artificial Intelligence Organization (IJCAI) 2022
pdf
Reinforcement Learning for Unified Allocation and Patrolling in Signaling Games with Uncertainty
, Elizabeth Bondi, Harshavardhan Kamarthi, Keval Dholakia, Balaraman Ravindran and Milind Tambe
International Conference on Autonomous Agents and Multiagent Systems (AAMAS) 2021
pdf
Real-time vessel segmentation and reconstruction for virtual fixtures for an active handheld microneurosurgical instrument
, Sara Moccia, Arpita Routray, Simone Foti, Elena De Momi, Cameron N. Riviere
International Journal of Computer Assisted Radiology and Surgery (IJCARS) 2022
pdf
Reviewing