About
I am an AI student at FPT University in Ho Chi Minh City, in my capstone year. I spend most of my time on computer vision, agents and on getting models out of notebooks and into something a person can click.
At FPT Software, I built AI that helps Honda catch battery defects. I designed a RAG chatbot that lets quality engineers ask questions about failures in plain language, and paired it with a vision-language model that reads defect images and explains what went wrong. I also worked on automotive quality analytics, turning messy production data into signals engineers could act on.
In my spare time, I build agents. My team's agentic RAG system took second place at the SEAL Hackathon Spring 2026.
What I care about most is the last mile: getting models out of notebooks and into the hands of people who use them. I’m looking for an AI Engineer role on a team that ships.
Based in Ho Chi Minh City · lenguyenquocanh.work@gmail.com
Off the keyboard
- Embedding models into single-board computers, for fun
- Watching NHRL robot fights every weekend
- Playing Valorant as 知己知彼 百战不殆#MK001
- Ping pong, still a newbie
- Big fan of Malcolm Todd
Psst. The last line is tuned to something: highlight it, a letter at a time.
How I run an ML project
- 01
Define the problem
Who is the user, what decision does the model inform, and what metric would they actually feel. I write this down before touching data.
- 02
Data
Collect, label, and look at it by eye. Measure label noise and class balance early, because they decide the ceiling.
- 03
Baseline
The simplest thing that could work, with a proper evaluation split. Every later number is judged against this.
- 04
Iterate on the model
One change per experiment, tracked in Weights & Biases. Keep the ablations, even the failed ones.
- 05
Evaluate
Held-out set, error analysis by slice, and a latency budget on the target hardware.
- 06
Ship
Export, containerise, add a demo and a health check. If no one can run it, it is not done.
Skills
Numbers in brackets are projects on this site that use the skill. Click one to see them.
Languages
Vision & modelling
Multi-view & simulation
LLM & agents
- LangChain (1)
- LangGraph (1)
- OpenAI API (2)
- RAG (1)
- Azure OpenAI
- Azure Cognitive Search
Deployment & data
- FastAPI (3)
- Docker (3)
- PostgreSQL (2)
- Supabase (2)
- Redis (1)
- Dataiku
- Power BI
- Git
Web, mobile & realtime
Education
- Bachelor of Artificial Intelligence
- FPT University, Ho Chi Minh City, September 2023 to present. GPA 3.5/4. Capstone: multi-view pedestrian detection with partial supervision (group GFA26AI02, supervised by Dr. Dao Duy Son).
- Experience
- FPT Software, Quy Nhon: AI Engineer (Nov–Dec 2025) on a RAG/LLM chatbot for Honda battery quality with Azure OpenAI and Azure Cognitive Search; Data Analyst Intern (Sep–Dec 2025) on QAI Honda Vision automotive quality analytics with Python, R and Dataiku.
Certificates and publications
- Dataiku Core Designer Certificate Dataiku Academy
- Dataiku Advanced Designer Certificate Dataiku Academy
- Dataiku ML Practitioner Certificate Dataiku Academy
- Dataiku Generative AI Practitioner Certificate Dataiku Academy
- Dataiku Developer Certificate Dataiku Academy
- Dataiku MLOps Practitioner Certificate Dataiku Academy
- Machine Learning with Python IBM / Coursera
- Databases and SQL for Data Science with Python IBM / Coursera