At Vingroup's AI Talent Training Program, I lead a four-member team developing BRIDGE, a deployed research assistant for citation-grounded literature reviews and code reproducibility assessment. I designed its LangGraph workflow with human approval checkpoints, resumable execution, and retrieval from Semantic Scholar, OpenAlex, and arXiv.
We benchmarked BRIDGE against Ai2's Asta Paper Finder, achieving higher recall on RealScholarQuery and ScholarQuest. I also evaluated its reproducibility audit on held-out ICLR 2026 repositories, reaching SM@10 of 0.70 and EM@10 of 0.50 at under USD 0.03 per audit.
At AI VIET NAM, I conduct empirical research in Vietnamese VQA and NLP and supervise a 12-member medical image analysis research group. My first-author paper on Vietnamese VQA was accepted at AICI 2026; its difficulty-aware reasoning approach reduced inference time by 75% against the high-compute baseline.
I also founded PawCal, an AI veterinary automation platform, and built its Zalo OA chatbot for appointment scheduling, reminders, and customer enquiries. For my FPT University capstone, S.M.I.L.E, I owned appointment booking, payments, and identity verification across a microservice backend.
