I've built machine learning and AI products spanning clinical analytics, drone fleet optimization, deepfake detection, and retrieval-augmented search.
For an arrhythmia classification project, I applied SVM and Random Forest models across 5,000+ ECG samples and built a Streamlit application for automated visual reports and real-time cardiovascular risk estimates. In Aerovane, I developed regression and classification models for drone selection, delivery allocation, and route planning, improving operational matching efficiency by 24%.
I also engineered VeriPixel, a MobileNetV2-based deepfake image detection pipeline trained on 179,000+ images that achieved 83.25% classification accuracy.
For CrawlAI, I architected a RAG pipeline for crawling, extracting, embedding, and searching web content, using FastAPI, Streamlit, ChromaDB, LangChain, MCP, Sentence-Transformers, LLaMA 3.3, AWS EC2, and Amazon S3.
