○ Examined the applicability of using 'iterative' principal component analysis to derive state space models given error-corrupted input and output measurements○ Applying the above techniques to flight data provided by Dassault Systèmes○ Exploring the integration of topological techniques such as Tucker's Lemma and Borsuk Ulma theorem to refine and improve optimization strategies and data analysis methods

Guide : Dr. Shankar Narasimhan, M.S. Ananth Institute Chair Professor, IIT Madras○ Conducted an in-depth analysis of POD for dimensionality reduction of complex high-dimensional data○ Explored the implementation of Reduced-Order Models in high-dimensional reactor simulations for solving complex spatiotemporal profiles, including temperature and concentration distributions○ Aimed at enhancing the adaptability of reduced-order models (ROMs) derived from Proper Orthogonal Decomposition (POD) in response to changing system parameters

Improved recommendation systems and conducted in-depth behavioral analytics○ Designed a Bayesian probability and holo-entropy hybrid model on 2.2M records to identify unauthorized users○ Implemented a Matrix Factorization model with LightFM to suggest products, achieving a precision at 50 of 0.994○ Utilized EDA and ECOD to analyze extensive data patterns and gain valuable insights into customer behavior

○ Suggested strategies and equipment enhancements for expected maintenance cost reduction of 20%○ Conducted scale-up computations, enabling a seamless shift from the lab scale to plant production○ Implemented analytical techniques such as XRD and Gas Chromatography for precise purity analysis○ Received a full-time job offer in recognition of outstanding performance

Physics-Informed Neural Networks○ Performed Literature Survey of 20+ research papers to develop a fundamental understanding of PINNs○ Implemented concepts from 12+ papers using Python frameworks, namely PyTorch and TensorFlow○ Analysed the scope of PINNs, Deep-O-Nets, and Finite Element Methods to solve design problemsTraffic Bot○ Implemented an automated system to track vehicles and assist in traffic management optimization○ Utilized the YOLOv8 algorithm within a real-time vehicle tracking system to accurately detect and track vehicles○ Deployed the model on a Raspberry Pi v3 that gave an accuracy of 85% with an inference speed of 25 frames/second