I built a personal AI assistant with corrective RAG that routes requests across calculator, web-search, and retrieval tools, then checks relevance and verifies answers. I deployed it on AWS with FastAPI streaming, Docker, and fallback routing across Groq models.
As a Machine Learning Intern at FlyRank AI, I developed a binary classifier to flag content pages losing search visibility and used its predicted probabilities to support a refresh-review queue. I evaluated it with Precision@50, grouped client splits, and leakage controls, then documented the workflow and its limitations in a capstone paper.
