Mélifly: Radar Hive Monitor Against Asian Hornets (melifly.store)
Co-founded hardware startup project: a solar-powered box on the beehive with a 24 GHz Doppler radar that tells hornets (~100 wingbeats/s) from bees (~230) and counts traffic, a camera estimating hornet flight directions to triangulate nests, ESP32-S3 and 4G LTE-M, plus a web platform for hive health and nest maps. Pilot season with beekeepers in 2027.
Startup · IoT · Signal Processing
Station-Level Bike-Sharing Demand Prediction
Machine learning comparison (Random Forest vs Poisson regression) for short-term demand forecasting using weather, temporal, and urban features. Balances predictive accuracy with operational interpretability for rebalancing decisions.
Data Science
Modular AI Agent for Financial Time Series Analysis
LLM-powered agent (Ollama) for OHLCV analysis combining natural language interaction with deterministic, auditable computation of technical indicators. Solves reproducibility challenges in LLM-based quantitative finance.
Technical Report
QR Code Scanner Web App
Progressive Web App for scanning QR codes and opening authorized subdomains. Built with JavaScript and ZXing library.
PWA
Natural Selection Simulator: Evolutionary Optimization
Visual evolutionary algorithm simulator demonstrating natural selection principles. Population evolves to optimize a parameter through selection, crossover, and mutation. Built with Flask, Chart.js, and vanilla JavaScript.
Interactive