Portfolio

All Projects

Professional work and academic research — from production AI systems at enterprise scale to machine learning coursework at HCMUT.

Industry Work

Professional Projects

Production systems built at FPT Software AI Center in collaboration with global partners and research institutions.

Agentic AI Nov 2025 – Present

Healthcare Agentic Chatbot & Management Portal

🏢 FPT Software AI Center 👥 Team of 10 💼 Associate AI Engineer
  • Architected and deployed an Agentic Chatbot specialising in Healthcare Products using AWS Bedrock AgentCore — including Product Recommendation Agent, Health & Lifestyle Agent, and Product Education Agent
  • Engineered comprehensive Management Portal covering component configuration & setup, real-time monitoring dashboards, automated testing, and audit logging
  • Led technical transition from Proof of Concept (POC) to production-ready system; achieved system accuracy >90%
  • Optimised agent design for maximum component reusability and minimal latency
🏆 "Best Team" Award — FPT Americas (ST25)
AWS Bedrock AgentCore Agentic AI LLMs AWS Python
🎯 Recommendation Oct – Nov 2025

Smart Product Recommendation System

🏢 FPT Software AI Center 👥 Team of 10 💼 Associate AI Engineer
  • Developed a Smart Recommendation System for a skin health mobile app — using customer data, scan scores, health goals, product catalog, and market rules to generate personalised AI-driven suggestions
  • Ensured legal compliance with enterprise and regulatory requirements
  • Monitored latency and quality metrics; implemented quality gates and automated testing for production deployment; verified system accuracy >90%
Recommendation Engine LLMs Python AWS
☁️ Azure Aug – Oct 2025

Blog System with AI Search & Recommendation

🏢 FPT Software AI Center 👥 Team of 4 💼 AI Engineer Intern
Dr. Nguyen Duy Khuong (Principal Data Scientist), FPT Software AI Center
  • Designed and implemented a full-stack blog system with AI-powered semantic search and personalised recommendation capabilities
  • Integrated Azure AI Search, Azure Cosmos DB (NoSQL), Azure Cache for Redis, and Azure OpenAI (AI Foundry) for optimal performance
  • Built recommendation engine and low-latency caching strategy for personalised content discovery
Azure AI Search Azure OpenAI Cosmos DB Redis Python
📄 RAG Aug – Oct 2025

M3ARAG — Multi-Agent RAG for Document Understanding

🏢 FPT Software AI Center 👥 Team of 2 💼 AI Engineer Intern
Dr. Nguyen Duy Khuong (Principal Data Scientist), FPT Software AI Center
  • Designed and implemented a Multi-Agent RAG system for document intelligence supporting PDFs, HTML, Office documents, and plain text
  • Enabled locally deployable solution with GPU acceleration via Docling — zero cloud dependency, on-premises ready
  • Architected the multi-agent pipeline combining semantic search and generation capabilities
RAG Multi-Agent Docling GPU LLMs
🔬 Research Jun – Jul 2025

Dual Attention Model for Innovation Discovery

🏢 FPT Software AI Center 👥 Team of 2 💼 Research Assistant Intern
Prof. Kazuyuki Motohashi (The University of Tokyo) & Dr. Nguyen Duy Khuong (FPT Software AI Center)
  • Developed the Dual Attention Model — an attention-based deep learning model to extract technical and firm-related keywords from company websites, improving signal extraction for downstream innovation analysis
  • Implemented a Transformation Matrix to align Company and Patent datasets for cross-domain analysis
  • Executed end-to-end Innovation Discovery pipeline: web scraping, preprocessing, normalisation, training, evaluation, and reporting
Attention Mechanism Deep Learning PyTorch NLP Python
🔭 Research Jun – Dec 2024

Unsupervised Knowledge Graph Construction Framework

🏢 HCMUT 💼 Research Assistant
Mr. Bui Cong Tuan (PhD Candidate), HCMUT
  • Researched and designed an unsupervised framework for building Knowledge Graphs from scratch with minimal domain expert involvement
  • Analysed SOTA LLM performance in generating Entities and Intents from raw Vietnamese documents
  • Investigated SOTA techniques in Embedding, Dimensionality Reduction, and Clustering to reduce bias in abstract Knowledge Graph entities
Knowledge Graph LLMs NLP Clustering Python
Academic Work

Research & Coursework Projects

ML/DL projects from Machine Learning and NLP courses at HCMUT.

A01

Sentiment Analysis with Various Models

April 2025 · Team of 5 · Machine Learning Course

Implemented and compared multiple ML/DL models — Decision Trees, Naïve Bayes, SVM, XGBoost, Random Forest, MLP, and Bi-LSTM. Focused on feature transformation, high-dimensional data handling, hyperparameter tuning, and systematic model evaluation across architectures.

Python Scikit-learn Keras PyTorch Bi-LSTM
A02

Fine-tuning Language Models for NLP Tasks

April 2025 · Team of 5 · NLP Course

Fine-tuned pre-trained language models for sentiment analysis, question answering, and machine translation. Implemented and compared T5-Base, BART-Base, and Flan-T5-Small — assessed accuracy, efficiency, and compute trade-offs with full GPU pipeline on Kaggle and Colab.

HuggingFace T5 BART Flan-T5 PyTorch
A03

Detect AI-generated Text

December 2024 · Team of 3 · Programming Integration Course

Evaluated SVM, Random Forest, XGBoost, Neural Networks, and Distilled BERT for AI text classification in the educational domain. Analysed model performance detecting LLM-generated essays — focused on accuracy, generalisation, and robustness across text styles.

BERT SVM XGBoost Scikit-learn PyTorch