research Enhancing Protein Fitness with Deep Learning Dual Degree Thesis on sequence-structure fusion using LMs and GNNs for function prediction and generative sequence design Deep Learning for Enzyme Activity Prediction Sequence-to-function learning pipeline using protein language models Graph Neural Networks for Protein Inverse Folding Improving ProteinMPNN through data augmentation and pre-training strategies Revisiting RVL-CDIP Fixing label errors and train-test overlap in document classification benchmark Automated PII De-Identification for Document Datasets Large-scale pipeline for detecting and removing sensitive personal information from 400,000+ documents with synthetic data replacement Neural Collapse in Hierarchical Loss Functions Investigating neural collapse phenomena and hierarchical loss functions in deep classification networks fun Optimized Implementation of GPT-2 Enhanced GPT-2 with modern attention mechanisms and efficient training Robotic Personal Assistant Autonomous robot with computer vision, SLAM, and conversational AI capabilities hackathon AI for Green Finance 3rd place solution at Temenos Encryptcon hackathon evaluating climate projects
