At NOUS, I built an AI system that reads and checks payment requests, including scanned and photographed invoices. It has processed 1,500 requests and reads them correctly up to 99.1% of the time.
I also built and operate NOUS’s document Q&A assistant, which uses RAG and serves more than 5,000 people. An automated ingestion pipeline keeps its knowledge base current with hundreds of new policy documents each month.
For NOUS’s in-house ERP, I lead architecture and integrations across 10 modules and 220 MVP functions, working with department heads to define processes. The system is designed to replace Base and Haravan and connect with services including MISA, Viettel Post, Shopee, and TikTok Shop.
In my solo Zalo AI Challenge 2026 entry, I built an offline agent for Vietnamese business questions that scored 0.99 penalized accuracy on the public test. Earlier, I delivered Tableau reporting for EVN and built automation and game-simulation tools in Python, Node.js, and C++.

