Research
Manuscript (in preparation)
Computational modeling of electrophysiology recordings can predict octopus arm movement, Nitish Gedela, Sachin Salim, Julianna Richie, Autumn Mclane Svoboda, Cynthia Chestek, Anne Draelos, Galit Pelled, 2023
Labs
Neuroscience
Draelos Lab, University of Michigan May 2023 – Present
Dr. Anne Draelos, Assistant Professor, BME
Led a project analysing real-time octopus arm motion using deep learning (DeepLabCut) & unsupervised (ProSVD) methods to comprehend its adaptive functionality for rehabilitation technologies
Discovered significant statistical variations in kinematic features to stimulations across different arm locations
Extended the results to other species and presented a one-hour talk in the Neural Network journal club
Preparing publication; my codebase is now used by researchers studying other animals such as monkeys and pigs, demonstrating the broader impact
Cortical Neural Prosthetics Lab, University of Michigan Jan 2023 – Apr 2023
Dr. Cynthia Chestek, Associate Professor, BME, EECS, Robotics
Report
Developed a real-time finger kinematics prediction model using reinforcement learning tools (Gym, RLlib-Ray). Mentored by Joseph Costello.
Trained the RL decoder using Proximal Policy Optimization achieving success in more than 99% of the simulated episodes
Finetuned a feed-forward neural network that decodes the neural signals from motor cortex of non-human primates
Movement Control Lab, Indian Institute of Science, Bengaluru, India Oct 2020 - Dec 2020
Dr. Aditya Murthy, Professor, Centre For Neuroscience
Code
Researched on the role of internal fast feedback controls in human hand movements by studying the inter-trial variability of duration and peak velocity of hand movements made with two different amplitudes. Mentored by Dr. Varsha Vasudevan.
Analyzed hand trajectory variance revealing the role of internal fast feedback in dynamically adjusting hand velocity to halt at the target position
Computer Science
Computational Economics Lab, Indian Institute of Technology Kanpur, India Jan 2018 - Apr 2018
Dr. Swaprava Nath, Assistant Professor, CSE
Report | Presentation
Explored incentivizing shared rides through a randomized polynomial algorithm employing minimum spanning trees to determine the optimal route
Analysis showcased a 125% environmental enhancement in simulations
Posters & Presentations
Neuroscience
• Project Presentation: Modeling APL-Mediated Local Inhibition in the Fruit Fly Mushroom Body Slides Dec 2023
• Journal Club Presentation: Real-time behavioral analysis on octopus arm using transfer learning and stable streaming dimension reduction, Neural Networks Workshop, University of Michigan Slides Oct 2023
• Project Presentation: Effect of hub network connectivity on network firing patterns, Computational Neuroscience Slides Oct 2023
• Paper Review: Structure of the Zebrafish Locomotor Repertoire Revealed with Unsupervised Behavioral Clustering
Slides Aug 2023
• Paper Review: Deep learning-based behavioral analysis is capable of outperforming commercial solutions Slides Jun 2023
• Paper Review: Task Learning via Internally Rewarded Reinforcement Learning Based Brain Machine Interfaces Slides Apr 2023
Computer Vision
• Poster: Tom & Jerry in Real life - Translating Cartoon to Natural Images using Stable Diffusion Poster Dec 2023
• Paper Review: Attentional Constellation Nets for Few-shot Learning
Slides Nov 2023
• Project Presentation: Brain Tumor Segmentation using an ensemble of 3D U-Nets, Graduate Computer Vision Slides Dec 2022
Others
• Thesis Presentation: A Quantitative Comparison of Solo and Shared Ride, Game Theory Slides Apr 2018
Projects
University of Michigan
Modeling APL-Mediated Local Inhibition in the Fruit Fly Mushroom Body Sep 2023 – Dec 2023
Dr. Victoria Booth (MATH 568: Computational Neuroscience)
Report | Presentation
Demonstrated that local inhibition regulates sparsity of Kenyon Cell (KC) outputs comparably to global inhibition
Substantiated through simulations that local-random PN-KC connectivity enhances odor recognition accuracy
Translating Cartoon to Natural Images using Stable Diffusion Oct 2023 – Dec 2023
Dr. Stella Yu (EECS 542: Advanced Computer Vision)
Report | Poster
Trained a latent diffusion model to unconditionally generate images of both domains
Used a pre-trained image captioning model (BLIP) as a guidance to condition the diffusion generation
Parkinson’s Disease Progression Prediction Feb 2023 - Apr 2023
Dr. Chris Teplovs (SI 618: Data Manipulation & Analysis)
Report
Utilized machine learning regression models for Parkinson’s disease progression prediction using protein and peptide data
Contributed to improving the accuracy and objectivity of Parkinson’s disease diagnosis, enabling early detection and personalized treatment planning
Seizure Detection and Closed-Loop Control Mar 2023 - Apr 2023
Dr. Cynthia Chestek (BME 517: Neural Engineering)
Report
Researched closed-loop seizure mitigation using machine learning algorithms and control schemes
Explored mathematical frameworks to understand the dynamics of seizures and developed control techniques using Simulink
Brain Tumor Segmentation using an ensemble of 3D U-Nets Oct 2022 - Dec 2022
Dr. Andrew Owens (EECS 504: Graduate Computer Vision)
Report | Presentation
Implemented 3D U-net, a deep convolutional neural network, to segment subregions of brain tumor
Created an ensemble of multiple models trained with different hyper-parameters to reduce random errors
Predicted the whole tumor region with a high dice score of 80.5%
IIT Kanpur
Breaking a Visual CAPTCHA Jan 2017 - Apr 2017
Dr. Vinay P Namboothiri (Visual Recognition)
Computed shape contexts, the histograms that accurately describe the shape of each potential letterusing Canny edge detection method
Ran affine-transformations to transfer the input image to roughly align the shapes with the matched templates
Solved EZ-Gimpy CAPTCHA by running affine-transformations of shape contexts and comparing them with pre-processed templates
Quora Question Duplication Feb 2018 - Apr 2018
Dr. Harish Karnick (Natural Language Processing)
Report
Implemented Time Distributed Layer along with GloVe embeddings achieving 83.9% accuracy in detecting semantic similarity
Demonstrated the significance of GloVe embeddings and attention-based techniques for improved NLP performance
Online Marketplace for Marginal Farmers Aug 2017 - Nov 2017
Dr. Swaprava Nath (Algorithmic Game Theory)
Report
Set up a theoretical online marketplace for the sale of farm-products to tackle the intermediary induced price inflation in the supply chain from rural farmers to urban consumers.