Dat Van

I am seeking a position in Artificial Intelligence where I can learn and grow across areas such as NLP, Computer Vision, and other AI-related domains. I aim to better understand how AI works at a fundamental level and gain hands-on experience applying it to real-world problems, while contributing effectively to the team.

AI Engineer at EasyFlight (2025-07 – Present)

Developed a conversational AI flight assistant with Langchain multi-agent system for natural language flight searches and real-time lookups.

• Developed a conversational AI flight assistant that enables users to search for flights using natural language, powered by a Langchain multi-agent system to handle both dialogue and real-time flight lookups.
• Optimized context window limitations by implementing a data extraction mechanism that parses massive JSON payloads from the flight API into concise summaries before passing them to the LLM.
• Implemented graceful error handling by converting strict API validation errors into natural language responses
• Applied strict prompting techniques and output formatting constraints to ground model responses in retrieved data, preventing hallucinations and ensuring seamless dialogue

AI Engineer at PIT Assistant (2025-10 – Present)

Built a Personal Income Tax chatbot using Knowledge Graph (GraphRAG) to retrieve context and provide answers with source citations for legal tax regulations.

• Built a Personal Income Tax chatbot that queries a Knowledge Graph (GraphRAG) to retrieve relevant context and provide answers with source citations, adapting to the July 2026 regulation updates.
• Formulated structure-aware document chunking tailored specifically for the hierarchical format of legal texts.
• Implement an LLM-powered extraction pipeline to identify entities and relationships from legal documents. Enforcing strict structured output constraints to ensure robust graph construction.

AI Engineer at YT translater (2025-12 – Present)

Developed a Chrome Extension with Flask backend for LLM-powered subtitle translation on YouTube with token-aware chunking and auto-retry mechanisms.

• Developed a Chrome Extension combined with a Flask backend to provide LLM-powered subtitle translation directly overlaid on the YouTube player.
• Designed a token-aware chunking pipeline to process subtitle segments in parallel, maintaining optimal context sizes for the LLM.
• Engineered a strict 1-to-1 line mapping system using numbered prompts, accompanied by an auto-retry mechanism that corrects the LLM when output lines are misaligned with the input.

Bachelor of Computer Science in Computer Science – University of Information Technology - VNUHCM (2022 – 2026)