Biomedical Sentence Similarity Tool

Semantic similarity in biomedical texts using neural network models with BioSentVec embeddings. Achieved 78.87% accuracy by minimizing mean squared error and evaluating semantic distances with cosine similarity and word mover's distance. Streamlined deployment with Docker and FastAPI for efficient integration.

Sentiment Analysis

Sentiment analysis on the 35 Million Amazon Reviews using an end-to-end MLOps approach. It classifies reviews as positive or negative, leveraging a robust pipeline that includes FastAPI for serving, Docker for containerization, and MLflow for model tracking.

LLM Fine Tuning

Fine-tuned a distilbert-base-uncased model for multi-class emotion classification (joy, sadness, love, anger, fear, surprise) using the emotion dataset. Implemented efficient training on an Apple M1 chip, completing in under 23 minutes. Applied to applications like sentiment analysis, chatbots, mental health assessments.

Natural Language Processing