About AuroraLens
Built to Answer One Question.
Aurora forecasting apps have existed for years. Most of them show you a global KP number and call it a forecast. I wanted something different — a tool that could tell me specifically, for where I am standing right now, whether the sky would light up tonight.
That question became AuroraLens.
Open Data. Real ML.
AuroraLens is built entirely on publicly available NASA and NOAA datasets. The DSCOVR satellite at the L1 Lagrange point — 1.5 million kilometers from Earth — continuously streams solar wind data. We process that telemetry through a dual-stage XGBoost model trained on 40 years of geomagnetic records from the NASA OMNI dataset, combining it with your local cloud cover to generate a 0–100 personal visibility score.
No proprietary black boxes. No fabricated accuracy claims. Just open science, engineered carefully.
Mosin Mushtaq
B.Tech Artificial Intelligence & Machine Learning.
Tech enthusiast passionate about space weather, exploration, and missions.
AuroraLens is my exploration of what happens when applied ML meets real-time geophysical data.
A dedicated space weather monitoring platform focused on infrastructure risk and solar event detection is currently in development.
NASA DSCOVR
Real-time solar wind telemetry from the L1 point. Updated every 60 seconds.
NASA OMNI2
40 years of hourly geomagnetic records used to train the prediction model.
Open-Meteo
Hyper-local cloud cover data by coordinate for visibility scoring.