Uncertainty Quantification in Vision Language Models. (Ongoing)(LLM), VLM, Retrieval Augmented Generation (RAG)Offline Policy Evaluation in Multi-Dynamic EnvironmentsMulti-Task Learning for Grape Cold-Hardiness PredictionProbabilistic ForecastingSim-to-real Learning of Footstep Constrained Bipedal Locomotion.AI Safety via Assistance GamesTeaching Assistant : ME 430 Systems Dynamics and Control

Working in the Technology Innovation Team.. Bayesian Optimization for Material Design - Discovering new recipes for synthesising better materials using Bayesian optimization with ~30 high dimensional training points. Properties of proposed recipes improve over recipes obtained through trial and error conducted for ~2years. · Multi Object Tracking - Using Kalman filters for state estimation and Hungarian algorithm for data association.· Deep Learning for Gaze Estimation· Edge Deployment of Deep Learning Models - Deploying pytorch models on Android by converting to ONNX and using OpenCV as a backend for inference.

Worked at Adaptive Computing Lab under Prof David Hsu· Implementation of a feature rich visualization tool based on Python Tkinter to visualize and debug QMDPNet, a deep learning algorithm for solving POMDPs.· Robot infrastructure setup for executing actions output by QMDPnet reliably and safely. Implementation of a robust position controller on the Fetch robot.· Imitation Learning for autonomous driving in an unstructured environment.

Worked on developing a robust system for traffic sign detection, recognition and tracking.

Worked on developing a low cost LiDAR system using a Teraranger One ToF sensor on a pan tilt unit for 3D scanning.· Experimented with various depth perception techniques such as structured light, stereo, ToF for implementing obstacle detection for a visually challenged person.

Worked at Robotics Research Center under Prof K Madhava KrishnaResearch into an End-to-end learning based approach for visual servoing in diverse scenes.Implementation of ’Guess from Far Recognise when Near’, a system for object search in unknown environments via frontier based navigation, far object recognition using 2D image segmentation and near object recognition using a bag of words model trained on 3D point clouds.· Deep Learning for Table Interest Point Detection - Research to find interest points or corner points of tables ina scene using cues from semantic segmentation and vanishing lines. Availability of semantic information such as interest points can help mobile robots navigate in a better way.· Automating GrabCut for Multilabel Image Segmentation - Implementing multi label Image Segmentation without user guidance by learning a Gaussian mixture model for each label and performing alpha expansion algorithm using MRF2.2 Library.

Applied Decision Trees and Support Vector Machines and other classification algorithms for classifying mutations as cancerous.