Open to opportunities

Minh Chí Nguyễn

@minhchnguyn

Message

I build multilingual RAG and agentic AI systems for document intelligence.

Vietnam

Message

What I'm looking for

I'm looking to build production-grade multilingual NLP, RAG, and agentic AI systems, especially for document intelligence, legal reasoning, and enterprise knowledge retrieval.

I'm contributing to PolicyVerse at VinUniversity and VinNLP, developing multi-agent systems for multilingual policy reasoning over Vietnamese legal text. I build agentic RAG pipelines and verification mechanisms for citation grounding, contradiction detection, and statutory interpretation.

Previously, I built production knowledge-retrieval and intelligent document-processing solutions at IMT Solutions, including an IDP pipeline handling 2,000+ invoices daily with over 95% field-level accuracy. My work spans LLM fine-tuning, multilingual retrieval, reranking, OCR, vision-language models, FastAPI, and scalable ML services.

Experience

Work history, roles, and key accomplishments

Current

Research Assistant

Current

VinUniversity

May 2026 - Present (3 months)

Contributing to PolicyVerse, an international research project on multilingual policy reasoning, developing agentic RAG pipelines and verification mechanisms for Vietnamese legal text.

RAG Agentic AI Contradiction Detection

AI Engineer Trainee

VinBigData

Jul 2025 - May 2026 (10 months)

Earned placement in the competitive AI Engineer Training Program (top 8% of 900+ applicants). Worked on production conversational AI and multi-agent orchestration frameworks.

Conversational AI Multi Agent Systems Orchestration

AI Engineer

IMT Solutions

Jun 2024 - Jul 2025 (1 year 1 month)

Developed production AI solutions for enterprise knowledge retrieval and intelligent document processing. Built an internal RAG system and deployed scalable AI services using FastAPI and Docker.

RAG fastAPI Docker Vision Language Models Production AI Systems

Data Scientist Intern

FPT Software

Mar 2024 - Jun 2024 (3 months)

Fine-tuned encoder-only transformer models (BERT, RoBERTa) on Vietnamese datasets for comment classification. Built an end-to-end text classification pipeline.

BERT RoBERTa Fine Tuning Feature Engineering Model Evaluation