Finalist — Top 24 / 200 · Techcombank HackCX-Together 2025
Published:
Placed in the Top 24 out of 200 teams at Techcombank HackCX-Together 2025, building an AI-powered banking assistant for enhanced customer experience.
Published:
Placed in the Top 24 out of 200 teams at Techcombank HackCX-Together 2025, building an AI-powered banking assistant for enhanced customer experience.
Published:
Placed in the Top 2 out of 300 teams at GDGoC Hackathon Vietnam 2026 in the AI Content Architecture track, building InsightForge — a multi-agent content automation system.
Published:
An asynchronous, agentic RAG pipeline for news retrieval and summarization, featuring two-stage retrieval with Milvus + BGE-Reranker and a fine-tuned Qwen2.5-1.5B served via vLLM.
Published:
A sophisticated multi-agent system using Stateful LangGraph and MCP servers to automate trend discovery, script generation, and cross-platform publishing for content creators.
Published in Symposium on Information and Communication Technology (SoICT) 2025, 2025
We investigate multi-modal large language models with a two-stage fine-tuning strategy for robust fashion image retrieval, demonstrating strong performance across standard benchmarks.
Published in ACM International Conference on Multimedia Retrieval (ICMR) 2026, 2026
We introduce VietFashion, the first benchmark for sketch–text composed image retrieval targeting culturally-specific Vietnamese traditional outfits, enabling fine-grained cross-modal retrieval via domain-adapted encoder fine-tuning.