About Skills Experience Projects Education Contact
Available for opportunities

PHAN VAN BANG

AI Engineer

Specializing in Multi-Agent Systems, RAG Pipelines, Voice AI & IoT Integration

Scroll
01.

About Me

AI Engineer with 2+ years of professional experience and 56+ AI projects spanning Multi-Agent Systems, RAG pipelines, Voice AI, IoT, Computer Vision, Trading, and Social Media Analytics. Built production-grade AI platforms including a multi-agent orchestration system with 94+ MCP tools, Swarm Intelligence, and continual learning capabilities across 8+ business domains.

2+
Years Experience
94+
MCP Tools Built
500+
Daily Conversations
56+
AI Projects

Short-term Goal

Contribute as an AI Engineer at a product-driven company, applying expertise in Multi-Agent Systems and RAG pipelines to build scalable AI solutions that directly impact business metrics.

Long-term Goal

Evolve into a Senior AI Engineer / Tech Lead role, driving AI architecture decisions, mentoring team members, and contributing to open-source AI communities.

02.

Technical Skills

Languages

Python TypeScript/JavaScript C# Java Go Rust C++ C

AI & LLM

LLM Context Engineering (Hermes) Fine-tuning (LoRA, QLoRA) RAG & GraphRAG Multi-Agent Orchestration Swarm Intelligence ReAct Reasoning Loop Deep Analysis Engine Self-Reflection & Critic Agent Autonomous Goal Engine Smart Model Routing Dreaming Engine Tree Search Planning Ontology Auto-Generation Continual Learning (XSkill) MCP Protocol (94+ tools) ComfyUI Workflows HuggingFace vLLM

Voice AI

OpenAI Whisper STT TTS (OpenAI/Fish Audio/ElevenLabs) VAD Full-Duplex Voice Chat

Computer Vision

Face Similarity API PaddleOCR-VL SAM3 Stable Diffusion ComfyUI Face Fusion Image Generation Pipelines

NLP

spaCy (Vietnamese + English) BGE Embeddings Tesseract Vietnamese OCR Text Chunking Strategies

Frameworks

LangChain LlamaIndex FastAPI CrewAI AutoGen Unsloth TorchRL llama-cpp-python Streamlit React React Native Next.js Express.js Spring Boot Laravel

Databases

PostgreSQL + pgvector Vector DB (Qdrant, Faiss, Chroma) MySQL SQLite MongoDB Firebase Neo4j (Graph DB) UMAP Dimensionality Reduction

Architecture

Microservices RESTful API WebSocket SSE Streaming Docker Compose Celery + Redis MCP Protocol (Client + Server + OAuth 2.1 PKCE) MQTT (IoT) OpenAI-Compatible API

DevOps & Cloud

AWS (S3, EC2) Docker Hub Cloudflare GitHub Actions Vercel VPS Management Nginx RunPod (GPU Serverless) Let's Encrypt SSL Multi-stage Docker Builds

AI IDEs & Dev Tools

Cursor Claude Code Codex Roo Code (Custom Modes)

Tools

Git N8N Automation Data Crawling/Scraping Playwright Unit Testing PaddleOCR Bonito (Data Synthesis) PyInstaller Binance API
03.

Work Experience

AI Engineer

MekongAI

06/2023 - Present
  • Built and maintained 56+ AI projects spanning 8+ business domains (Legal, Education, Recruitment, POS, Healthcare, Crypto, Social Media, Content Generation).
  • Architected and deployed a Multi-Agent Orchestration System with 94+ MCP tools, supporting dynamic function routing, Swarm Intelligence for collaborative problem-solving, and XSkill continual learning - processing 500+ daily conversations with 85%+ autonomous resolution rate.
  • Developed MekongAI Nexus core agent system with ReAct reasoning loop, Deep Analysis Engine, Self-Reflection, Critic Agent, Autonomous Goal Engine, and Tree Search Planning.
  • Implemented Dreaming Engine with 3-phase background memory promotion (Light → Deep → REM sleep cycles) for continual knowledge consolidation.
  • Designed and built a production RAG pipeline with hybrid search (semantic + keyword + reranking), GraphRAG for relational knowledge, and multi-modal document processing - achieving 92% answer accuracy across 10,000+ knowledge documents.
  • Developed Full-Duplex Voice AI system integrating Whisper STT, custom TTS (Fish Audio/ElevenLabs), and VAD with <500ms latency - deployed on IoT devices (ESP32) for real-time conversational AI.
  • Built face similarity API and computer vision pipelines (SAM3, PaddleOCR-VL on RunPod GPU serverless) for production image analysis.
  • Created real-time crypto monitoring system with MA crossover signals and Telegram alerts (Binance API integration) for automated trading signals.
  • Developed AI recruitment platform with CV parsing, Vietnamese NLP (spaCy), and job-candidate matching algorithms using vector similarity search.
  • Fine-tuned Llama 3.1 models (8B/70B) using Unsloth framework with trajectory compression → SFT data pipeline for domain-specific AI agents.
  • Built AI-powered learning platform with automated content generation using ComfyUI workflows and LLM pipelines - producing 1000+ quiz questions and educational content items with human-level quality.
  • Implemented microservices architecture using FastAPI, Docker Compose, Celery + Redis for async task processing, WebSocket for real-time communication, and SSE for streaming responses - supporting 100+ concurrent users.
  • Managed full DevOps lifecycle: AWS deployment (EC2, S3), Docker containerization, CI/CD via GitHub Actions, Nginx reverse proxy, and VPS management for 99.5% uptime across all services.
04.

Key Projects

01
Multi-Agent System

MekongAI Nexus

Multi-Agent Orchestration System

Production multi-agent platform with 94+ MCP tools, Swarm Intelligence, XSkill continual learning, dynamic routing, and context-aware orchestration handling 500+ daily conversations.

Python FastAPI LangChain MCP Redis Docker
02
AI Content Platform

MekongAI Knowlympic

AI Learning Content Platform

Automated educational content generation using ComfyUI workflows and LLM pipelines, producing 1000+ quiz questions with RAG-powered knowledge retrieval and quality validation.

Python ComfyUI RAG LLM PostgreSQL
03
AI Personality

MekongAI MetaHuman

AI Personality Platform

AI personality system with customizable personas, voice cloning, multi-modal interactions, and context-persistent memory for natural, human-like conversational experiences.

Python TTS LLM WebSocket React
04
IoT + AI

ESP32 AI Bot

IoT Voice Assistant

Hardware voice assistant on ESP32 with real-time STT/TTS, LLM integration via WebSocket, VAD for natural conversation flow, and <500ms end-to-end latency.

ESP32 Whisper WebSocket FastAPI VAD
05
AI Recruitment

AI Recruitment Platform

AI-Powered Recruitment with NLP Matching

AI-powered recruitment with CV parsing, Vietnamese NLP matching, vector similarity search for job-candidate matching, and spaCy Vietnamese + English dual NLP pipeline.

FastAPI Qdrant spaCy OCR AWS S3
06
Multi-Agent Content

AI Content Generator

Multi-Platform Content Generation

AI Agents auto-generate content for Facebook, Instagram, TikTok with RAG-based brand voice learning. Multi-agent pipeline (Research → Write → Edit → Publish).

CrewAI Streamlit Qdrant RAG Multi-LLM
07
Video/Photo Analysis

MekongAI TikVault

AI Video/Photo Analysis Platform

AI-powered video/photo analysis platform using Google Gemini. Content categorization, trend detection, and engagement prediction for social media optimization.

FastAPI React 19 MongoDB Gemini AI
08
Crypto Trading

Real-time Crypto Monitor

Automated Trading Signal System

Real-time cryptocurrency monitoring with technical analysis (MA crossover signals), automated Telegram alerts for trading signals, and multi-exchange data aggregation.

Streamlit Binance API Telegram Bot TA
05.

Education

B.Eng. in Information Technology

Vietnam-Korea University of Information and Communication Technology (VKU)

Da Nang, Vietnam

Relevant Coursework:

Machine Learning Deep Learning Natural Language Processing Software Engineering
2020 - 2025
06.

Certifications & Languages

Languages

Vietnamese Native
English Professional Working

Certifications

Deep Learning Specialization concepts
LangChain & LLM Application Development
LLM Fine-tuning (Unsloth, LoRA/QLoRA)
Multi-Agent Systems Architecture (production-deployed)
07.

Soft Skills

Self-learning

Rapid adaptation

Technical Communication

Clear documentation

Problem Decomposition

Analytical thinking

Team Collaboration

Technical leadership

Ownership Mindset

Deadline-driven execution

Multi-domain AI Expertise

8+ industries covered

End-to-end Product Dev

Concept → Production

08.

Get In Touch

I'm currently open to new opportunities where I can contribute my expertise in AI engineering. Whether you have a question or just want to connect - feel free to reach out.