LOADING...

0%

Projects

Hands-on builds with real impact

NexAI Edge-to-Shelf: Farm AIoT Platform
2025 – 2026
AWS • GCP • Gemini 2.5 • RAG • LINE Bot • Firebase
  • Built Farm2Market AI – live product "Dr. Chang's Happy Farm" (changfarm.netlify.app) – a LINE-integrated agentic platform reducing agricultural order error rates by 15% and saving farmers 10–15 hours of manual reconciliation weekly across 68 verified retail channels
  • Engineered multi-modal AI agents using Gemini 2.5 with agentic workflows, RAG via Gemini File Search API, and AWS/GCP/Firebase deployment; kept AI processing costs below 1% of transaction revenue
  • Finalist, 2026 Best AI Awards (MOEA, Taiwan) and Startup Homecoming+ recognition from Taiwan's Ministry of Digital Affairs; core engine covered by a US Provisional Patent
WindGuard MLP: AI-Powered Fault Prediction
Dec 2025
MLP • SHAP • React • Python • Tailwind CSS
  • Developed a predictive maintenance system using MLP achieving 98.1% accuracy in detecting gearbox and generator faults
  • Integrated SHAP analysis to visualize how parameters like oil temperature and rotor torque influence fault probability
  • Architected a real-time monitoring dashboard with React and Tailwind CSS featuring automated risk alerts
Carbon-Trace: IoT-Based Carbon Tracking
Sep 2025
Python • Blockchain • IoT • Encryption
  • Designed a decentralized IoT ecosystem leveraging blockchain immutability for tamper-proof carbon records – Runner Up, VIT YUVA Club Hackathon, top among 250+ participants
  • Built hardware sensor logic and a real-time dashboard, plus encrypted TLS-equivalent RESTful gateways connecting IoT sensors to the backend to prevent data manipulation in transit
  • Automated emission calculations and data normalization using AI agents for live sensor streams
Layer 7 DDoS Simulation & Mitigation
Mar 2024
Python • TCP/IP • Network Security • WAF
  • Built a multithreaded HTTP flood simulator exploiting TCP/IP application-layer behavior to identify server bottlenecks and stress-test network resilience
  • Engineered a software Firewall with Rate Limiting and behavioral traffic analysis (IDS-style pattern recognition) to detect and block Slowloris, GET/POST flood, and connection exhaustion attacks
  • Validated WAF rule configurations against real attack traffic logs, simulating IPS-level inspection to harden application defenses