MethaneGuard AI

MethaneGuard AI was inspired by the challenge of detecting methane leaks early and helping organizations decide where inspections should happen first. Methane is invisible and difficult to monitor continuously, while satellite and environmental datasets can contain large amounts of information that is difficult to turn into actionable decisions.

I built MethaneGuard AI as an AI-powered environmental risk monitoring prototype. The dashboard uses the NASA/ORNL DAAC AVIRIS-NG methane and greenhouse-gas plume dataset as its real-world reference and presents monitoring information through an interactive Streamlit dashboard. It displays methane levels, leak probability, risk scores, severity, risk levels, and inspection priorities for monitored sites.

The prototype was developed in Python and deployed using Streamlit Community Cloud, with the source code maintained on GitHub. The main challenge was creating a practical end-to-end solution using limited computing resources while keeping the prototype simple enough to demonstrate clearly. The current risk values are simulated for system validation, while the NASA dataset provides the real-world reference for the solution.

The goal is to develop this prototype further into a more advanced detection and inspection-prioritization system using real hyperspectral imagery, machine-learning models, geospatial analysis, and validated methane plume detection.

What I learned

This project taught me how to turn an environmental problem into an end-to-end data and AI prototype. I learned how to structure risk indicators, build an interactive Streamlit dashboard, work with a real-world NASA reference dataset, deploy an application from GitHub, and communicate technical results through a simple user interface. I also learned the importance of clearly distinguishing simulated prototype outputs from validated real-world model predictions.