SafeWing AI

On June 12, 2025, Air India Flight 171 crashed in Ahmedabad just 40 seconds after takeoff, killing 280 people. We made a dashboard to allows pilots to inspect potential failures before it's too late.

Comment

Here's a concise hackathon project description for SafeWing AI:

Inspiration

Air India Flight 171's tragic crash in June 2025 killed 280 people due to preventable system failures. We built SafeWing AI to predict potential aircraft subsystem failures to allow pilots to inspect before they become catastrophic.

What it does

Real-time aviation safety dashboard with 3D aircraft visualization that monitors 6 critical subsystems (engine, hydraulic, electrical, control surface, cabin, altimeter) using LSTM neural networks and random forest to predict Remaining Useful Life (RUL) and prevent crashes.

How we built it

  • Frontend: Next.js with Three.js for 3D aircraft visualization
  • ML Backend: LSTM and random forest models deployed using FastAPI, Docker, and GCP
  • Data: NASA CMAPSS turbofan engine dataset + synthetic subsystem data
  • Integration: Real-time sensor monitoring with color-coded alerts on actual aircraft parts

Challenges we ran into

  • CORS issues with external ML API requiring proxy implementation
  • Complex 3D model integration and part highlighting
  • Real-time data synchronization between multiple subsystems
  • Matching API data formats with frontend expectations

Accomplishments that we're proud of

  • Successfully integrated live LSTM predictions with 3D visualization
  • Created realistic takeoff simulation with authentic sensor degradation
  • Built comprehensive multi-system monitoring (engines + 5 subsystems)
  • Achieved sub-100ms API response times

What we learned

  • Real-time ML integration challenges in web applications
  • Three.js 3D visualization and material manipulation
  • Aviation system complexities and failure patterns
  • Importance of predictive maintenance in safety-critical systems

What's next for SafeWing AI

  • Integration with real aircraft sensor data streams
  • Advanced anomaly detection algorithms
  • Regulatory compliance for aviation safety standards
  • Partnership with airlines for pilot training simulations

Built With

Share this project:

Updates