AeroFail

Why I built a Mechanical Failure Predictor out of a plane Crash:
When Jeju Air Flight 2216 went down on December 29, 2024, killing 187 people, I did what a lot of people did, which was that I watched the news, felt awful, and moved on with my day.
Mechanical failures rarely come from nowhere. They come from signals that were there and weren’t caught or were caught and weren’t acted on. 187 people boarded a flight that data might have flagged. I wanted to understand that chain well enough to ask better questions about it.

So I built AeroFail.

It’s a web platform where engineers and/or students learning to think like engineers can input real sensor parameters from a flight system and get a predictive failure risk score across six subsystems: engine, fuel system, hydraulics, structural, avionics, and landing gear. There’s a chatbot that explains what the model found. A simulator that runs stress analysis on aircraft components. And a section documenting real accidents and what failed, why, and what a model like this might have caught. The model is imperfect. The physics are simplified. The data is synthetic for now. But the questions are real. And that feels like a place to start.