Things I've built
At SEL, I build the testing and AI infrastructure that keeps power grids reliable. Outside of work, I like to explore ideas through side projects — from open-source MCP servers to deep-learning stock models, hackathon builds to a full AgriTech platform serving Vietnamese farmers.
Each project below started with a question I couldn't stop thinking about. Scroll down to explore.
10+
Projects shipped
40+
GitHub stars
2
Award-winning
4
Categories
All Projects
Filter by category to explore what I've been working on.
An MCP server that integrates Wireshark/tshark with AI tools and IDEs. Capture live traffic, parse .pcap files, apply display filters, follow streams, and export JSON through Claude Desktop, VS Code, or CLI.
Deep learning-powered stock guidance system using a Multi-Modal Late Fusion architecture: a stacked LSTM over 60-day OHLCV + technical indicators fused with FinBERT sentiment from financial news.
High-concurrency Go CLI that fires one prompt at multiple LLMs simultaneously, streams results in parallel via SSE, and renders a live BubbleTea dashboard showing each model racing to the finish.
Physics-aware supply chain control plane using NVIDIA Cosmos World Models and a Multi-Agent Council. Reduces Time-to-Action for disruptions from 48 hours to 300 seconds.
Statistical evaluation harness analyzing LLM token entropy and log-probabilities to detect silent model uncertainty during insecure code generation. Includes A/B testing, CWE-based test cases, and comparative visualizations.
Ensemble machine learning pipeline (Random Forest, CART) evaluating 48 distinct features to classify phishing websites with 98% accuracy. Awarded Best Project at UIC Engineering Expo.
AI-powered hacker onboarding platform that analyzes GitHub repos and generates personalized contribution roadmaps using Gemini, MongoDB Atlas, and optional voice mentorship.
FarmLink
AgriTech platform backed by ADC Vietnam connecting 200+ farmers and buyers across 4 provinces. Distributed AWS infrastructure with 99.99% uptime. Accepted into NVIDIA Inception & Google Cloud programs.
MCP server exposing AWS S3 tools (list buckets/objects, presigned GET/PUT), safe by default. Designed for Claude and other AI agents.