Skip to content
21.03°N 105.85°EHanoi, VietnamOpen to roles · Hanoi or remote

Le Van Trung TinhSoftware Engineer.AI Product Builder.

Building full-stack applications, AI-powered products, and practical software solutions.

I focus on the parts users never see but always feel — APIs, data, background jobs — and increasingly on the AI features built on top of them.

fig. 01 — System map
TipThe shape of the systems I build. Hover or tap a node to see the stack and the projects where I used it.
Focus
Full-stack & AI-powered products
Stack
TypeScript · Java · PostgreSQL
Now
Product engineering at Margick
Before
Backend intern at Tomosia

01/ 07About

Engineering ideas into real products.

Based in
Hanoi, Vietnam · GMT+7
Education
B.IT, Duy Tan University, 2025
Builds with
TypeScript, Java, PostgreSQL
AI in use
Gemini, OpenAI, Claude Code
Code
github.com/tinhne

I'm Tinh, a software engineer based in Hanoi. I studied Information Technology at Duy Tan University in Da Nang, interned on production backend services at Tomosia, and now work on product engineering at Margick, a startup building a SaaS website builder.

Most of my work sits where backend systems meet the product: APIs, data models, background jobs, and the interface that makes them useful. Lately that includes AI — putting models like Gemini and OpenAI behind real features, and using coding agents to move faster without lowering the bar on review and testing.

My engineering approach

  1. 01

    Understand

    Start from the user and the constraint, not the stack. Agree on what “done” means before writing code.

  2. 02

    Design

    Sketch the data model and the boundaries first — what must be consistent, what can run in the background.

  3. 03

    Build

    Small, typed, reviewable increments. AI agents speed up the typing; the decisions stay mine.

  4. 04

    Verify

    Reproduce a bug before fixing it. Test the paths that can lose data, money or trust.

  5. 05

    Improve

    Ship, watch how it's actually used, then fix what hurts most — not what's most fun.

Currently exploring

Areas I'm actively learning. Interested, practising — not claiming expertise.

  • AI application engineeringAgent workflows, tool calling, evaluating output quality
  • LLM-powered productsRetrieval, structured output, cost and latency trade-offs
  • Backend architectureEvent-driven design, queues, idempotent jobs
  • Software qualityTesting strategy, observability, readable code at scale
  • AutomationConnecting APIs into workflows that remove manual steps

02/ 07Experience

Where I've done the work. Real teams, real codebases.

  1. Present

    Startup · Vietnam

    Margick (opens in a new tab)

    Software Engineer · Product Engineering

    Margick builds a multi-tenant SaaS website builder with AI-assisted site creation, on top of the WordPress / WooCommerce ecosystem.

    • Build and maintain website features and templates inside the multi-tenant website builder.
    • Turn Figma designs into responsive, production-ready pages across desktop, tablet and mobile.
    • Use AI coding agents (Claude Code) daily for implementation, refactoring, debugging and documentation — every change still goes through my own review.
    • Debug site functionality and fix layout, performance and quality issues before release.

    Stack

    WordPressWooCommerceElementorPHPJavaScriptClaude Code
  2. Mar 2025 — Oct 2025

    Internship · Da Nang

    Tomosia Vietnam (opens in a new tab)

    Backend Developer Intern

    Six months on a product team building backend services that ran in production.

    • Built and maintained backend services with Node.js (Fastify) and TypeScript, structured with Clean Architecture.
    • Designed REST and tRPC APIs, paying attention to query efficiency and consistent error handling.
    • Containerized services with Docker and wrote Bash scripts to automate deployment steps.
    • Contributed Next.js and shadcn/ui components to internal tools.

    Stack

    TypeScriptNode.jsFastifytRPCDockerBashNext.js
  3. Jan 2025 — May 2025

    Part-time · Da Nang

    Hoc Cung Chuyen Gia (opens in a new tab)

    Power BI Tutor · Part-time

    One-to-one tutoring in data visualization and reporting.

    • Taught data modeling, DAX and dashboard design to individual learners.
    • Learned to explain technical decisions clearly to people without an engineering background.

    Stack

    Power BIDAXData modeling

The one-page version, with education and certifications, is in my resume (opens in a new tab).

03/ 07Projects

Selected work. Explained end to end.

P—012026Source available

RetailOps Intelligence

An operations inbox for online stores: tested rules find the problems, an AI copilot explains them.

A B2B SaaS that imports a store's products, orders and stock from CSV, detects oversold SKUs, late shipments, refunds owed and inconsistent data with nine deterministic rules, and gives the team a prioritised inbox where every fix is verified by the next data import.

Role
Solo developer — product discovery, domain design, full stack, AI, testing, Docker and CI
Focus
Domain modelling and grounded AI
Stack
FastAPIPostgreSQLNext.jsTypeScriptPlaywrightDockerGemini
P—022025Live demo

Meet.AI

Video meetings where an AI agent joins the call, then writes the notes.

A full-stack SaaS where users create AI agents, meet with them over video, and get a transcript, a structured summary and a chat that remembers the meeting.

Role
Solo developer — design, full stack, deployment
Focus
Event-driven AI processing
Stack
Next.jsTypeScripttRPCDrizzleInngestGeminiOpenAI
P—032026Live demo

Book Review Platform

A Spring Boot API with JWT auth, roles and a React client.

Readers browse and search a catalogue, write reviews and like others'; admins manage books, categories, users and reviews. Backend-first, with a typed React client.

Role
Solo developer — API, data model, client
Focus
Auth, transactions, API design
Stack
Java 21Spring BootSpring SecurityJWTPostgreSQLReact
P—042026Source available

Vehicle Telemetry System

10 Hz sensor streaming to an Android Automotive dashboard.

A Spring Boot backend streams simulated vehicle telemetry over WebSocket to an Android Automotive app, raises threshold warnings, and stores history for a REST API.

Role
Solo developer — backend, Android client, infrastructure
Focus
Real-time data, throughput, resilience
Stack
Java 21Spring Boot 3WebSocketPostgreSQLDockerKotlin

Archive

Earlier builds, smaller scope
  • 2025

    Devhelp — Mezon developer bot

    Bot for saving commands, logging bugs by severity and linking fixes, with full-text search and soft delete. Built for the Mezon Bot Challenge.

    NestJSTypeScriptSQLiteMezon SDK
  • 2025

    Medical diagnosis chatbot

    Capstone: symptom-to-diagnosis engine combining XGBoost with association rules, trained on cleaned data mapped from the Disease Ontology, behind Spring Boot auth.

    PythonXGBoostAprioriPandasSpring Boot
  • 2025

    Blog API

    Blog backend with JWT auth and Redis caching, structured with Clean Architecture.

    Node.jsFastifyTypeScriptRedis
  • 2024

    Clinic booking system

    Appointment scheduling, patient records, JWT auth and video calls between staff and patients.

    Node.jsExpressReactPostgreSQLStream
  • 2023

    MazaBank — desktop banking

    JavaFX app with admin and client roles, transfers, deposits and account limits, persisted in SQLite.

    JavaJavaFXSQLite

04/ 07AI & Product Building

Building with AI. Engineering beyond prompts.

I use AI in two places: inside products, as features people actually touch, and inside my workflow, as a tool that makes me faster. In both, the design, the review and the testing stay my responsibility.

To be precise about it: I'm an application engineer, not an ML researcher. I don't train models — I build reliable software around them.

fig. 04 — How I work with coding agentsThe agent writes drafts. I own the outcome.
  1. 01Frame

    Me
    Requirements, constraints and what “done” means
  2. 02Plan

    Me
    Choose the approach and the boundaries
    Agent
    Explore the codebase, surface options
  3. 03Implement

    Agent
    Draft code in small, reviewable diffs
  4. 04Review

    Me
    Read every diff; reject what I can't explain
  5. 05Test

    Me
    Verify behaviour and edge cases
    Agent
    Scaffold tests, reproduce bugs
  6. 06Ship

    Me
    Release, observe, decide what's next

Capabilities, with receipts

  • Professional work
  • Shipped in a project
  • Exploring

A

AI-powered product development

Putting models behind features users actually touch.

  • Gemini API

    Shipped in a project

    Gemini 2.5 Flash generates structured meeting summaries via Inngest Agent Kit.

    → Meet.AI

  • OpenAI Realtime API

    Shipped in a project

    Connected through Stream so an agent can speak inside a live video call.

    → Meet.AI

  • Background AI pipelines

    Shipped in a project

    Slow model calls run as retryable job steps, never inside a request.

    → Meet.AI

  • Rapid prototyping

    Shipped in a project

    Getting from idea to a working, deployed version quickly enough to learn from it.

    → Meet.AIDevhelp Bot (opens in a new tab)

B

AI-assisted engineering

Coding agents as part of a disciplined workflow, not a replacement for one.

  • Claude Code

    Professional work

    Daily driver for implementation, refactoring and codebase exploration.

    → Margick

  • Debugging with agents

    Professional work

    Reproduce first, then let the agent trace and propose — I confirm the root cause.

    → Margick

  • Docs & test scaffolding

    Professional work

    Agents draft documentation and test skeletons; I decide what's worth asserting.

    → Margick

  • Multi-agent workflows

    Exploring

    Splitting larger tasks across agents with clear briefs and review gates.

C

Product automation

Removing manual steps by wiring systems together.

  • Event-driven workflows

    Shipped in a project

    Webhooks feed Inngest functions that update state and notify users.

    → Meet.AI

  • Deployment scripting

    Professional work

    Bash and Docker to automate repetitive release steps.

    → Tomosia

  • Chat-ops bots

    Shipped in a project

    A team bot that turns chat commands into searchable records.

    → Devhelp Bot (opens in a new tab)

  • Data processing

    Shipped in a project

    Crawling, cleaning and mapping source data into a usable training set.

    → Medical Chatbot (opens in a new tab)

D

Technical product thinking

Deciding what to build, and how much engineering it deserves.

  • Requirements analysis

    Professional work

    Turning a vague request into acceptance criteria before touching code.

    → Margick

  • Architecture decisions

    Shipped in a project

    e.g. broadcasting at 10 Hz but persisting at 1 Hz to cut writes tenfold.

    → Vehicle Telemetry

  • Feature development

    Professional work

    Shipping inside an existing product and codebase, not just greenfield.

    → MargickTomosia

  • Iterative improvement

    Professional work

    Testing with real usage, then fixing what hurts most.

    → Margick

05/ 07Skills

Tools I've shipped with. Each one traced back to real work.

Backend11

Frontend7

Data & infra11

AI & product6

Also worked with5

Listed tools are ones I've used in professional work or shipped personal projects. Things I'm only learning live under Currently exploring.

06/ 07Achievements & Education

Milestones. Competitions, certificates and the degree behind them.

Education

Duy Tan University

Bachelor of Information Technology

Aug 2021 — Jun 2025 · Da Nang, Vietnam

  • Competition

    LG Dream Code — LG

    Top 30 · Finalist

  • Competition

    Mezon Bot Challenge — Mezon

    Contestant · HelpDev Bot · Mar — Apr 2025

    Demo (opens in a new tab)
  • Certificate

    Cybersecurity Fundamentals — AUGEST

    Completed · Sep 2024

    Certificate (opens in a new tab)
  • Workshop

    Build with AI Workshop — GDG Central

    Participant

07/ 07Contact

Let's build something real.

Hiring for a software, full-stack or AI product role — or building something that needs another engineer? Email is the fastest way to reach me.