Aspiring AI Solution Architect & Data Automation Builder.
Having transitioned from a solid foundation in Business (BTEC FPT, 2024) to aggressive Data Trading and Quantitative execution, I am now fully committed to building leverage through AI-driven Automation Systems. My long-term goal is to evolve into a full-stack AI Solution Architect—solving complex workflow problems and generating hyper-efficient revenue streams through programmatic automation.
Currently extending my formal CS fundamentals at UIT (VNUHCM, enrolled Mar 2026, expected 2028) via a distance-learning program. This strategic choice allows me to operate full-time in the startup/hackathon ecosystem while actively riding the current AI wave.
"Outcome over Output. I engineer leverage for high-performance teams to scale efficiency without scaling headcount."
📍 Based in: Ho Chi Minh City, Vietnam
🎓 Education: UIT – VNUHCM (B.Sc. IT, Distance Learning, 2026–2028)
💼 Focus: AI Agent Architectures · Automation & Scraping · Quantitative Trading
🏆 Goal: AI Solution Architect → Building High-Leverage Autonomous Systems
Actively pushing the boundaries of what local AI running on Edge devices and web architectures can achieve in competitive hackathon environments:
| Project | Description | Tech |
|---|---|---|
| vietfi-advisor | AI Financial Assistant for Gen Z Vietnam (WDA 2026) — Full-stack system integrating LLM conversational engines with personal finance aggregation. | TS/JS, Next.js, AI Agents |
| antigravity-self-learning | 7 Self-Learning Models for AI Coding (Code Smell, ADR, Prompts) extracted from 14 real-world projects. | Local AI Pipelines |
Building "zero-human" autonomous systems that research, decide, and execute:
| Project | Description | Tech |
|---|---|---|
| social-lead-gen | Crawl (Playwright) → Classify (DeepSeek R1) → Auto-outreach pipeline. | Python, DeepSeek |
| telegram-channel-analyzer | Scrape & analyze Telegram channels - NLP keyword analysis + engagement tracking. | Python, NLP |
| business-deep-research | AI business research + Devil's Advocate debate (Gemini + Google Grounding). | Prompt Chaining |
Backtested, production-grade systems for Forex (XAUUSD), Crypto, and Prop Firm challenges:
| Project | Description | Tech |
|---|---|---|
| ex5-backtest | 🆕 MT5 Auto-Backtest Pipeline — Headless CLI: Compile→Test→Rank 70 EAs. | PowerShell, MQL5, CLI |
| xauusd-smc-signal-engine | AI Trading Engine — XGBoost + Smart Money Concepts. EV+0.183R, 78% monthly win rate | Python, XGBoost |
| btc-smc-rl-bot | Reinforcement Learning (PPO) learns BTC entry points, auto-exports MT5 EA | Python, Gymnasium |
| xauusd-ichimoku-rl-bot | RL (PPO) multi-timeframe Ichimoku for XAUUSD. +7.12% test, PF 3.76, WR 86.7% | Python, Gymnasium |
| comarai-algo-promax | SMC + Trend MAs + Proxy ML Triggers — Pine Script + Python backtesting pipeline | Pine Script, Python |
| mql5-multisignal-dca-ccbsn | 9 Signal Modes × DCA × Anti-Detect — MT5 EA with Python ML Optimizer | MQL5, Python |
| hcc-reader | Reverse-engineered MT5 .hcc/.hc binary format parser | Python |
📊 More Trading Tools & Indicators (click to expand)
- pinescript-ict — ICT SSL Premium & Discount PineScript v6 indicator
- indicator-tradingview — TradingView indicators collection
- moondev-agent — Trading AI Agents: Ichimoku + MQL5 Backtest Pipeline
- mql5-algo-trading-portfolio — Full MQL5 EA portfolio
- telegram-copy-signal — Signal copy dashboard
AI & Automation Pipelines
Data Science & Backend
Quantitative Tools
I do not write code line-by-line; I orchestrate swarms of Local and Cloud reasoning models to deploy production-grade software at extreme velocity. My daily workflow burns through tens of millions of tokens, managed through a custom proxy architecture.
- Routing & Resilience: Utilizing 9Router load-balancing across multi-platform LLM endpoints (Claude Code, Antigravity Ultra, Codex Plus) to bypass aggressive rate limits and maintain 24/7 uptime.
- Token Operations: Running RTK (Rust Token Killer) as a CLI proxy layer to aggressively strip payload noise, securing 60-90% token savings across continuous loops.
- State & Memory Management: Employing Superpower plugins to hot-swap high-context structural memory across concurrent ChatGPT instances during massive monolithic refactoring.
My orchestration bypasses basic "single-shot" prompting. By leveraging Harness Chain Thinking, I build looping, sequential pipelines where agents act as Devil's Advocates, validating and recursively auditing each other's outputs before compiling execution code.
Beyond code, I deploy custom Model Context Protocol (MCP) tool-chains. Similar to standard APIs, these MCPs fetch raw data, format executive layouts, and natively generate polished PowerPoint Presentations and Word Briefs—subsequently piping insights directly into Power BI for automated C-level dashboarding.
Currently actively researching Hermes Agent integrations to establish a fully autonomous headless agent capable of crawling repos, patching bugs, and resolving issues asynchronously without human observation.
University of Information Technology – VNUHCM (2026–2028) B.Sc. Information Technology (Distance Learning)
Learning remotely by choice to allocate full-time bandwidth toward building high-impact AI/Automation pipelines and competing in hackathons while completing formal CS fundamentals.
BTEC HND in Business (Graduated 2024) Foundation in business analysis and strategic thinking — Bridging the gap between code output and commercial value.
Open to exciting opportunities in AI Architecture, Data Automation, and Growth Systems.
| 📞 Contact | |
|---|---|
| hungphamphunguyen@gmail.com | |
| GitHub | github.com/hungpixi |
| Website | phamphunguyenhung.com |
Built with passion, caffeine, and AI logic.