Led research team of 5, funded by Swinburne to build a multi-agent system with domain-specialist agents (diagnostics, pharmacology, triage) coordinated by a reasoning agentic orchestrator MCP:

Led a self-sourced team (contracted via Teqar) to build a client–server, plugin-style multi-agent assistant that ingests PDFs/DOCX/XLSX, normalizes to Markdown and JSON objects, performs semantic chunking + RAG, and auto-completes technical outcome docs with grounded, auditable facts.

Our team of 4 believing in domain-specific, reliable and affordable education, we support university students and graduates in both their studies and career, leveraged by AI and technology capability.

Senior MCP architecturer, building local assistant that fuse LLM, NLP, OCR on the solution to turn all unstructured PDF documents into direct industrial AI-assistance that can provide guidance for businesses. Our partnered businesses including Nvidia, HPA, IBM, Juniper, and CISCO Vietnam.

Contributor of CLARA, Vietnamese nationally recognized medical AI assistant (LLM and VLM), trained on 400k+ doctor-annotated cases, including 210k+ X-rays, MRI scans, and synthesis / distillations.

Led team of 4 software to design and prototype an AI-powered predictive maintenance system for Sky Ledge fleet operations. Oversaw the full development lifecycle including external dataset collection, Python-based data pipelines, and model training using XGBoost, RandomForrest and LSTM. Coordinated frontend and backend development to deliver a functional dashboard for real-time vehicle health monitoring using OBD-II data. Managed task planning, research direction, and client engagement preparation.

Develop a comprehensive web-based student management platform designed for language schools to handle student enrollment, class management, placement testing, and administrative tasks. The system supports both Vietnamese and English interfaces and serves multiple user roles including students, staff, and administrators.

Leading a team of 6 to build a Vietnamese-language AI assistant, converts business questions into SQL queries across MySQL and Oracle, fusing multi-modal self-QA for Reinforcement Learning over failures and successes. Built with FastAPI, Gemini & Qwen3, MongoDB vectorbase, leveraging Chain-of-Thought and Agentic RAG for autonomous analytics in sales, supply chain, and marketing. The system reduces manual workforces by up to 80%, delivers instant, explainable insights, a unique plugin solution enables direct database-to-decision workflows without human intervention.

Led a 4-member team to architect and build a scalable electricity forecasting and control application from IoT device. Integrated data cleaning pipeline, including KNNImputer for data backfiller and abnormalities detection, with LinearRegression imputer for timestamp fallbacks. Integrated ML/DL (TFT, XGBoost), Reinforcement Learning (DQN), and LLM-driven advisory to predict billing, optimize device usage, and automate energy-saving decisions. Enabled multimodal interaction via ASR, text, and voice synthesis for a fully interactive LLMOps experience. Partnered with EVN Vietnam

Participate in fintech digitalization projects with full-stack development (Vert.x Backend, Java Android) and AI integration (CV, ML, OCR, LLM).

Analysing data, document, article for Sentiment Analysis and Business Overview with LLM models, integration with cloud computing architecture. Leading current team to leverage Obsei framework and Transformers for sentiment analysis, fact-checking, summarization, deliver actionable business insights for the clients.

Developed the official website for a local commuter café at Auburn Station. Features include interactive menu, real-time pre-ordering, loyalty program with personalize QR code, and staff dashboard for order management, rating and feedback, sale analysis. Built with VueJS, Node.js (Vercel), database storage on MongoDB, and Firebase. Revolutionizing the business with digitalisation procedure, skyrocketing sale on an instance by 30%. Find more here.

Android solution streamlining staff rostering and payment management with Kotlin, Node.js, and MariaDB.

Develop Water Meter OCR, a real-time water meter scanning solution using Next.js, Node.js, and Flask API, for government client. Self-mining and annotation of data, employs YOLOv5xu for detection and CV+OCR for text, hosting on Jetson Nano and AWS EC2. Production met the planned KPI to reduce manual labour by approximately 35%.

Dev-Ops Engineer, deploying software architecture using Docker. Deploying basic backend infrastructure with Docker and JavaScript.

Hardware technician specializing in motherboard replacement and repair services. Have an in-hand experience with routers motherboard.