I built a workflow of four specialized AI agents that finds promising product deals, estimates fair value using RAG and a fine-tuned LLM, and alerts users when it finds an opportunity.
I also built an AI knowledge assistant that makes internal company documents searchable through natural-language questions. It uses LangChain, ChromaDB, MiniLM embeddings, and Qwen 4B.
I fine-tuned Gemma 2B with QLoRA on 15K product examples to predict prices, with an average prediction error of about $10.25. I later integrated the model into an autonomous AI system for price estimation.
On a contract with Innolitics, a U.S.-based AI healthcare company, I worked on an AI medical imaging project involving CT scan alignment and anatomical segmentation using Python-based imaging tools. My other projects include medical image and text classification, heart disease prediction, and bulldozer price prediction.
