Machine-Generated Text Attribution

Trained BERT Sequence classifier on a novel dataset created using 5 generative models (GPT-2-small, GPT-2XL, Phi-2, Falcon-7B, Mistral-7B-it) by prompting with text sourced from Wikipedia and GSM8K Math dataset, achieving F1-score of 0.65. and studied the impact of input length, prompt domain, and LLM parameter size on attribution accuracy.

Tiktokenizer.js

Developed a static site in pure JavaScript for visualizing the GPT-2 Byte-Pair Encoding (BPE) tokenization process, replicating official OpenAI API

Federated Learning - MNIST

Implementated Federated Averaging (FedAvg) algorithm on MNIST dataset using TensorFlow and Keras

Brain Tumor Segmentation from MRI using PSPNet

Demonstrated efficacy of PSPNet for segmentation of brain tumors from MRI data, achieving Dice: 0.66 and IoU: 0.552