Full-stack × AI systems
Full-stack engineerputting LLMs intoreal products.
Scroll
Now
Full-stack engineer at Kite Metric, owning AI/LLM, fintech and enterprise engagements end to end. Going deeper on GraphRAG, self-hosted small language models and making LLM workflows reliable.
Selected work
01 — 4 projects · 2023 — now- 01
AI Interview Agent
A voice AI that runs first-round candidate interviews, evaluates them with LLMs and hands employers a structured report.
→ 100% of evaluations delivered after the async redesign
- OpenAI
- Gemini
- n8n
- Vapi
- React
- 02
AI Matching Businesses
An LLM-powered pipeline that turns messy business data into structured profiles, then matches and recommends partners — with every score explainable.
→ Every recommendation ships with the reasons behind it
- OpenAI
- Python
- PostgreSQL
- 03
Audit-Grade AI
The data pipeline behind an audit-AI platform — raw domain material in, task-specific datasets out, ready for LLM retrieval or small-language-model fine-tuning.
→ One pipeline, any dataset the next model needs
- Python
- OpenAI
- Gemini
- LlamaIndex
- Unsloth
- 04
Car Care System
A booking and garage-management platform for four very different users — customers, merchants, staff and admins — in one system.
- React
- NestJS
- Tailwind CSS
- Supabase
Product-minded engineering
I turn rough conversations intoproducts people can actually use —shaping the experience, architectureand the intelligence behind it.
- 3+
- years shipping production software
- 4
- clients worked with directly, requirements to release
- 3
- domains shipped in: AI/LLM, fintech and enterprise
- 3.6
- GPA · Excellent degree, VNU-HCM University of Science (2025)
Honors
- B.Sc. IT with Excellent degree — VNU-HCM University of Science2025
- Binh Dinh Fellowship Scholarship2022
- Ranked 1st, Faculty of IT — K20 cohort2020
How I work
03 — Process- 01
Listen
Sit with the client, map the real workflow, and write down what “done” means before any code exists.
- 02
Architect
Turn requirements into a schema, service boundaries and an honest plan for where the LLM belongs — and where it doesn't.
- 03
Build
Own the feature end to end: UI flows, API, database and the deployment that puts it in front of users.
- 04
Harden
Make it survive reality — async workflows instead of fragile webhooks, traceable AI decisions, boring deploys.
Experience
04 — Career / 2020—NowA career built by taking on more of the problem.
- Aug 2024 — PresentHo Chi Minh CityCurrent chapter
Kite Metric JSC
Full-Stack Engineer
- Core engineer across concurrent AI/LLM, fintech and enterprise engagements for 4 clients, mostly early-stage startups — each owned from architecture through implementation.
- Works directly with clients to gather requirements, propose technical solutions and design system architecture aligned with the business.
- Apr 2023 — Aug 2024Ho Chi Minh City
LEAP LLC
Full-Stack Engineer
- Built digital-transformation and B2B/B2C booking platforms for multiple clients, replacing manual workflows with web systems.
- Developed search, booking, ticketing and seller-management modules with Angular, Java Spring Boot, Docker and PostgreSQL.
- 2020 — 2025Ho Chi Minh City
University of Science — VNU-HCM
B.Sc. Information Technology · Excellent degree
- Graduated 2025 with an Excellent degree, overall GPA 3.6 / 4.0.
- Ranked 1st in the Faculty of IT, K20 cohort (2020). Binh Dinh Fellowship Scholarship (2022).
The toolbox
05 — Working set / 06 disciplinesTools change. The standard doesn't.
A practical stack for taking products from interface to infrastructure — chosen around the problem, not the trend.
- 01
Frontend
Interfaces & product flows
- React
- Next.js
- Angular
- Remix
- Tailwind CSS
- 02
Backend
Services & system boundaries
- Node.js
- NestJS
- Python
- FastAPI
- Java Spring Boot
- 03
AI / LLM
Retrieval, agents & evaluation
- OpenAI
- Gemini
- Anthropic
- LlamaIndex
- RAG
- GraphRAG
- Agents
- LLM workflow orchestration
- 04
Data
State, search & vector
- PostgreSQL
- MongoDB
- Supabase
- Qdrant
- 05
Mobile
Cross-platform delivery
- React Native
- Flutter
- 06
Cloud / DevOps
Ship, observe & scale
- GCP
- Azure
- Docker
- GitHub Actions
- Vercel
- Kubernetes
Writing
- 01
Webhooks time out. LLM workflows don't care.
Why long-running LLM evaluations broke behind synchronous webhooks, and the async pattern that fixed it.
- 02
A match nobody can explain is a match nobody trusts
Notes from building an LLM-powered business matching pipeline where traceability mattered as much as accuracy.
- 03
Designing one platform for four kinds of users
What building a garage booking system for customers, merchants, staff and admins taught me about schema-first product work.
