Building Agentic Systems from Scratch: Lessons from MeHubb
Why LLMs beyond chatbots require structured JSON planning, state machines, and autonomous tool calling to reliably execute code in production.
Read Article →
Available for AI Product & Systems Engineering
Kim Ngan Le Huynh
AI Product Engineer · Builder · Systems Thinker
"I design and build AI-powered systems that turn ideas into real, working products."
My background spans mobile engineering, AI/LLM integration, product development, QA, and automation. I enjoy working at the intersection of technology, product, and real-world human needs.
01 / ABOUT ME
Turning Ambiguous Problems Into Working Systems
Hi, I’m Kim Ngan. I’m a software engineer and product builder interested in how technology can turn messy, ambiguous problems into simple systems that actually work.
My background started in mobile development, working with iOS, Android, and Flutter. Over time, I moved closer to product management, automation, and AI — building systems that connect software workflows with real human needs.
Today, I focus on AI-powered products, agentic systems, and product engineering.
Problem
→
Product
→
Architecture
→
Prototype
→
Production
→
Iteration
* Rather than stopping at a prototype, I prefer taking systems all the way to production.
What I’m Interested In
Currently Exploring
- ▹ Autonomous AI agents & multi-tool execution
- ▹ LLM application architecture & prompt orchestration
- ▹ Advanced RAG & structured memory systems
- ▹ Self-healing build pipelines & workflow automation
- ▹ Product-led experimentation & AI-assisted development
How I Think & Work
Systems Thinking
I naturally look at workflows, dependencies, infrastructure, and the relationship between components rather than isolated tasks.
Product Thinking
I start with the core problem and user need before deciding what technology or framework to use.
Builder Mindset
I learn by building. When I need a new capability, I prefer creating a working prototype and learning from real constraints.
Cross-Discipline Integration
My strongest work sits at the intersection: AI + Software Engineering + Product Strategy + UX + Automation.
Documentation & Process Design
I turn complex technical work into clear workflows, frameworks, checklists, and reusable operational systems.
02 / FEATURED PROJECTS
Built Systems & Case Studies
I build small systems to explore ideas, solve real problems, and understand what happens when AI software meets real users.
MeHubb
AI App Builder · Agentic Systems · Self-Healing Workflow Automation
An agentic platform that turns natural-language ideas into working web applications. Instead of manually going through idea → specification → coding → build → deployment, MeHubb connects these steps into an integrated AI-driven system.
Agentic Execution Pipeline
01
User Request
Natural Language
→
02
LLM Brainstorm
Claude / GPT
→
03
JSON Plan
Structured Spec
→
04
Code Gen
PHP / JS / CSS
→
05
Self-Healing
Error Recovery
→
06
Deploy
Docker / VPS
What I Built
- LLM-Powered Planning: Autonomous PO/BA agent analyzing raw briefs into brand specs.
- Structured JSON Plans: Strict schema validation for execution steps.
- Code Generation Engine: Multi-file asset generation with full theme support.
- Docker & VPS Pipeline: Automated remote rsync, Nginx server block creation, and SSL Let's Encrypt.
- Telegram Bot Interaction: Full remote command menu for deployment & post management.
- Self-Healing Recovery: Automatic build error diagnosis and code self-repair loop.
Tech Stack
MePlann
AI Life Planner · Personal Productivity · BrainDump to Action Engine
An AI-powered productivity system designed to help people turn thoughts into structured actions. When people have too many thoughts, tasks, and ideas, the core issue is cognitive overload — MePlann transforms messy BrainDumps into prioritization & clear momentum.
Product Architecture & Data Flow
Input
BrainDump
Voice / Text
→
Interface
iOS / Telegram
SwiftUI / Bot
→
AI Engine
Task Structuring
Priority Scoring
→
Backend
Firebase / Memory
State Store
→
Output
Calendar & Action
Google Sync
Key Features & Responsibilities
- Natural-Language Task Input: Instant parsing of messy thoughts into structured tasks.
- Cognitive Score & AI Coaching: Dynamic popup banners providing gentle micro-break prompts.
- Cross-Platform Ecosystem: Native iOS application + Telegram Bot + Web Dashboard.
- Google Calendar & Reminders Sync: Two-way calendar synchronization.
- End-to-End Ownership: Designed product architecture, UI/UX, Swift codebase & AI integration.
Tech Stack
AI Experiments & Prototypes
Small experiments where I test ideas, architectures, workflows, and interactions with AI models.
01
Natural Language → Structured Task
How can an LLM turn an unstructured thought into an actionable, priority-scored task object with zero hallucination?
LLM Applications
02
AI Execution Planner
Converting high-level goals into machine-readable, multi-step JSON execution plans for autonomous tools.
Agentic Systems
03
Self-Healing Build Pipeline
Allowing AI agents to inspect runtime log output, diagnose compilation errors, and execute corrective code patches automatically.
Automation
04
Cognitive Load Reduction
Designing AI coaching interfaces that help users reduce mental overload without taking control away from them.
Human + AI
05
AI-Assisted Software Lifecycle
Accelerating the pipeline from idea to production by orchestrating LLM tool-calling and automated deployment.
Dev Tools
03 / TECHNICAL JOURNAL
Writing & Engineering Notes
Notes from building products, working with AI, learning software engineering, and thinking about how complex systems work.
Why Most Personal Productivity Apps Fail (And How MePlann Tackles Cognitive Overload)
Most tools focus on adding more features. But the real barrier is cognitive friction. Here is how we designed BrainDump to Action parsing.
Read Article →Designing Self-Healing AI Workflows: Beyond Prompt Engineering
What happens when a code generation attempt fails? Building closed-loop feedback mechanisms that let AI diagnose and patch its own build errors.
Read Article →From Mobile Developer to AI Product Engineer: 5+ Years in Software
How working across iOS, Android, QA analysis, and product management shaped my ability to build complete systems from 0 to production.
Read Article →
04 / EXPERIENCE
Career & Work History
5+ years of software engineering experience across mobile development, product analysis, quality assurance, and AI product building.
International & Remote Environment
Working across Flutter, iOS, and Android to deliver high-quality mobile products and AI-assisted engineering workflows.
- End-to-end mobile engineering and product feature delivery.
- Integrating AI-assisted development tools and LLM features.
- Cross-platform architecture and quality-aware software design.
MeHubb & MePlann Systems
Building independent AI products from initial problem discovery to production deployment.
- Designed agentic AI pipelines converting prompts to deployed Docker containers.
- Architected MePlann iOS application, Telegram bot integration, and Google Calendar sync.
Enterprise Products
Bridged engineering, product management, and quality assurance across mobile and web platforms.
- Automated mobile & API testing (Appium, Postman, JIRA).
- Requirement clarification, bug root-cause analysis, and cross-functional team leadership.
Financial & Business Applications
Built and shipped native iOS applications to the App Store.
- Swift, UIKit, SwiftUI, CoreData, REST API integration, and App Store submission.
Banking, IoT & Real Estate Systems
Engineered mobile applications connecting hardware protocols and real-time backend services.
- Kotlin, Swift, MQTT protocols, WebSockets, Firebase, Mapbox, and ARKit integration.
05 / CURRICULUM VITAE
Core Skills & Qualification
🤖 AI & LLMs
Claude API, OpenAI API, Prompt Engineering, AI Agents, RAG Concepts, Embeddings, Agentic Workflows.
💻 Engineering
Swift, SwiftUI, Flutter, Kotlin, Python, REST APIs, Firebase, Docker, VPS (Linux), Git, WP-CLI.
💡 Product & Systems
Product Discovery, User Stories, Requirements Engineering, Systems Architecture, Agile/Scrum.
🛡️ Quality Assurance
Appium, Postman, API Testing, JIRA, Test Automation, Root-Cause Analysis.
🎓 Education
Bachelor of Computer Software Engineering — Ton Duc Thang University (2017 – 2021)
06 / CONTACT
Let’s Build Something Together
I’m interested in AI products, product engineering, workflow automation, technical product consulting, and remote collaborations.
Email
huynhngan.58@gmail.com
LinkedIn
linkedin.com/in/nganhlk
Live Systems
Based In
Vietnam · Open to International & Remote Opportunities