At Wolters Kluwer, I architect production AI systems for Italian legal content, including RAG, document intelligence, and high-performance retrieval platforms.
I delivered an AI Legal Assistant spanning event-driven ingestion for 10M+ documents, hybrid retrieval over 30M vectors, and LangGraph agents with tool calling. It achieved 87% factual correctness, 82% answer relevancy, less than 2% hallucinations, and 93% recall@10.
I also replaced a legacy legal document analysis system with a Generative AI pipeline that improved accuracy from 60% to 98% and reduced operational costs by 40%. Earlier at John Deere, I deployed real-time construction-site obstacle detection models with F1 of 0.87 and sub-100ms edge latency.
Across these systems, I work with Python, AWS, FastAPI, LangChain, LangGraph, Transformers, Docker, Kubernetes, and NVIDIA Triton. I’ve led teams and projects, earned a U.S. patent, and teach 13,000+ paid students through Udemy courses.
