Explored novel encoding and prediction frameworks for the classification of umami peptides and molecules.Applied NLP-based encoders such as ProtBERT and Mol2Vec along with ML/DL predictors to encode and classify data, leading to over 90% accuracy

Implemented Ensemble learning methods to predict potential risk scores of severe illnesses based on the functional features and the abundance of species inside the gut microbiome, successfully getting recall greater than 75%.In collaboration with Dr. Himel Mallick at Weill Cornell Medicine, Transfer Learning is being used to predict metabolome functionality by analyzing species abundance.

Improved model interpretability and reduced computational load for datasets by optimising feature selection techniques.Reduced feature set from 5000+ to 80 while maintaining near-baseline performance by implementing and evaluating feature selection frameworks in supervised and unsupervised learning.

For DES102: Introduction to Human Computer Interaction, under Dr. Rajiv Ratn.For DES101: Design Drawing & Visualisation, under Prof Anoop Ratn.