At The Bank of New York Mellon (BNY), I refactored Log4j2 logging across 294+ files, separating logs into four dedicated loggers and reducing primary log-file size by 85%+ without data loss. I also automated logger discovery and resolved 324+ SonarQube issues.
At Indian Institute of Technology, Hyderabad, I independently implemented the KgCoOp vision-language prompt tuning architecture for federated learning. I redesigned training as a client-server pipeline, reaching 95.01% accuracy on IID data and 58.42% on non-IID data.
At Indian Institute of Science, Bangalore, I built an audio machine-learning pipeline for 250+ clinical voice recordings and fixed a data-corruption bug in label computation. My projects include SAARTHI, an emotion-aware recommendation system, and PETVET, which I co-developed for livestock disease prediction.
