↓ Skip to main content
Nguyen Phan Du | Software Architect & AI Agentic Specialist
Nguyen Phan Du

Nguyen Phan Du

Software Architect | AI Agentic Specialist | Foundation & Infra Engineer

Hello! I am Nguyen Phan Du (@badpaybad), a Senior Software Architect & Foundation Infra Engineer based in Hanoi, Vietnam.

With over a decade of hands-on engineering experience, author of 170+ open-source repositories, and recognized as an Arctic Code Vault Contributor, I specialize in designing robust software foundation infrastructure, architecting high-throughput distributed backends, and pioneering PC-based AI Agentic Workforce systems.

💎 Core Capabilities & Strategic Collaboration
#

Beyond writing code, I deliver end-to-end architectural solutions: Rock-solid Foundation – Deep Performance Optimization – Cost Efficiency – Exponential Speed via AI.

🏗️ 1. Foundation Infrastructure & High-Performance Architecture
#

For enterprises needing scalable architecture or solving critical performance bottlenecks.

  • Enterprise Architecture Patterns: Deep expertise in CQRS (Command Query Responsibility Segregation), Domain-Driven Design (DDD), Event Sourcing, Asynchronous Message Brokers (Redis, Message Queues), and Microservices.
  • Deep Low-Level Optimization: Eliminating communication overhead by compiling Native C/C++ shared libraries (.so/.dll) and invoking them directly via Dart FFI or CLR Interop for high-throughput computing, graphics, and real-time streaming.
  • Ultra-Lean Runtime Infrastructure: Building ultra-lightweight backend runtimes (Dart Server, .NET Core Serverless, Micro-frameworks) delivering single-digit millisecond latency while slashing cloud server bills by up to 70%.

🚀 2. Zero-to-One Product Engineering (Technical Co-founding & Partnership)
#

For founders and teams seeking to turn visionary ideas into production-ready platforms at record speed.

  • Full-Spectrum Technical Leadership: Covering the entire product lifecycle from foundation infra, database design, API Gateways, and real-time communications (SSE, WebSockets, WebRTC) to modern cross-platform apps (Flutter Mobile & Desktop).
  • AI-Accelerated Delivery: Leveraging specialized AI Agentic workflows and custom IDE harnesses to dramatically compress MVP build times, shipping to production in weeks rather than months.
  • Architected for Scale: Building clean, maintainable foundations from day one so you never have to throw away code when user traffic surges.

🤖 3. Pragmatic AI Agentic Workforce & Local LLM Solutions
#

For organizations seeking genuine AI automation with zero cloud overhead and 100% data privacy.

  • PC AI Agentic Workforce (Local-first, No Cloud Server Required): Deploying autonomous 24/7 AI agents that run directly on your workstation or PC, controlled seamlessly via Telegram Bot or private webhooks without recurring cloud fees.
  • Skill-Based Harness & Token Optimization: Crafting project-specific rules and skill frameworks that activate the full multi-disciplinary power of AI IDEs while strictly minimizing token consumption.
  • On-Device & CPU-Optimized LLMs: Deploying and optimizing local large language models (llama.cpp, Vistral, Ollama) on CPU hardware for Named Entity Recognition (NER), local document parsing, and absolute privacy.

🌐 4. Technical Advisory & Global Engineering Network
#

For investors, startups seeking technical due diligence, or engineers wishing to connect.

  • Architecture auditing, technology stack evaluation, and tackling hard engineering challenges.
  • Technical exchange on computing evolution: Punched Cards $\to$ Machine Code $\to$ Assembly $\to$ High-Level Code (C#, C++, Dart) $\to$ AI Agentic Workforce.

⚡ Battle-Tested Proof of Work (170+ Repositories)
#

Highlights from open-source repositories on GitHub @badpaybad:

🤖 AI Agentic, Local LLMs & Computer Vision
#

  • a.i-assistant-chatbot-telegram-serverles: PC AI Agentic Workforce – Personal AI assistant running locally on PC, controlled via Telegram, zero cloud server needed.
  • Ner_Llm_Gpt: CPU-optimized execution of Vistral LLM via llama.cpp for Named Entity Recognition (NER).
  • InsightFace_TF & insightface: 2D/3D deep facial analysis and recognition pipelines on TensorFlow.
  • pixel_link & OcrTesseract: Instance segmentation for text detection in natural scenes and advanced OCR.
  • darknet: YOLOv4 real-time object detection neural network implementations.

⚡ Low-Level Native Interop & High-Performance Media
#

  • flutter-dart-ffi-dlib-opencv: Building native C/C++ (dlib, OpenCV) shared libraries (.so) invoked directly via Dart FFI on Android/Flutter with zero bridge latency.
  • rplayer: Ultra-low-latency RTSP video streaming player for Flutter using direct Dart FFI instead of slow MethodChannels.
  • cameraip2youtubelive: Ingesting IP camera RTSP streams via OpenCV and live-transcoding directly to YouTube via FFmpeg.
  • dlib-android & libuvc: Cross-platform USB video device control and embedded computer vision.

🏗️ Enterprise Foundation Architecture & Distributed Backend
#

  • cqrs-dot-net-core: Enterprise reference boilerplate implementing CQRS on .NET Core with Redis as a distributed message bus.
  • opendotnet-ecommerce-platform: Distributed eCommerce platform built on CQRS, DDD, Event Sourcing, and Microservices.
  • dart-run-as-httpserver-api-sse-ws: High-throughput Dart HTTP server supporting REST APIs, Server-Sent Events (SSE), and WebSockets.
  • mini-mvc: Lightweight C# WebHost runtime supporting HTTP & WebSockets without heavy framework overhead.
  • dotnetcore-serverless: Serverless boilerplate architected for AWS, Azure, and Google Cloud.

🛰️ Robotics, Real-time Streaming & Edge Computing
#

🎯 Why Partners & Clients Trust Me
#

  1. Hardware-to-Application Mastery: I don’t stop at framework wrappers; I operate comfortably across kernel levels, memory management, native C/C++ bindings, and distributed cloud layers.
  2. Long-Term Architectural Discipline: Every system is designed with clean boundaries, high observability, robust error handling, and complete documentation.
  3. Proven Real-World Execution: Not just theory — all architectural patterns are validated by 170+ repositories and deployed into live production environments.
  4. Pragmatic Cost-Efficiency: Always recommending the most economical and performant path (PC-based AI agents, token reduction strategies, lean server footprints).

📬 Get in Touch & Start a Project
#

Let’s discuss your next system architecture, product launch, or AI automation workflow: