Developed a custom Healthcare Correspondence Assistant using large language models to intelligently extract, filter, and summarize patient data from scanned records, reducing manual document review time by over 50% for hospitals and care providers. Processed and indexed more than 10,000 healthcare documents by integrating OCR pipelines, Chroma database, and Llama 2, enabling real-time retrieval of relevant data for clinicians and administrative staff through a streamlined query interface. Deployed and optimized the model using GPU-based inference on AWS, cutting average response times from 2.5 minutes to 15 seconds and improving query accuracy by 20%, while ensuring cost-efficient scalability for large healthcare systems.

Developed an ensemble of machine learning models (InceptionV3, ResNet, VGG) achieving 95.6% accuracy for MRI-based brain tumor detection. Developed and fine-tuned an ensemble model combining multiple classifiers, demonstrating the potential of ensemble techniques in enhancing diagnostic accuracy

Developed Python regression models for wine quality prediction, achieving 0.634 RMSE and identifying key quality drivers through analysis