Autonomous Predictive Maintainence for Vehicles

Inspiration Chaotic Indian streets, where roadside mechanics fix battered bikes daily, inspired Autonomous Fleet Brain. Reactive maintenance causes massive downtime—we built AI to predict failures like brake wear or engine overheat before they strand riders, blending agentic AI with telematics.

What it does Monitors fleets in real-time, detects issues via AI, generates personalized voice alerts, and visualizes risks on a 3D dashboard. "SCAN FLEET" triggers mobile scans → backend analysis → instant audio + color-coded vehicle updates.

How we built it Backend: Flask + Lang Chain/Deep Seek AI scripts alerts, gTTS generates audio, serves via /api/start-journey.

Frontend: Three.js 3D orbiting bikes, HTML/CSS dashboard with risk cards.

Integration: Macro Droid webhook → API → 3D/voice updates.

Challenges we ran into CORS blocked frontend-backend calls—fixed with Flask-CORS. Dynamic audio paths needed UUID filenames. Three.js lagged on mobile; optimized geometries. Spotty network broke Macro Droid webhooks—added error handling.

Accomplishments that we're proud of Live 3D fleet visualization with real-time AI alerts in under 48 hours. Seamless mobile-to-web-to-voice flow. Scalable from single bikes to delivery fleets, demo-ready in 30 seconds.

What we learned Lang Chain excels for urgent alert generation; Three.js shines for intuitive 3D UIs. Backend-frontend sync demands robust CORS/audio handling. Agentic systems scale predictive ML (XGBoost) to production effortlessly.

What's next for Autonomous Predictive Maintenance for Vehicles Real telematics integration (GPS/sensors), ML failure prediction models, multi-language alerts, cloud deployment (AWS), and fleet-scale dashboards for enterprise users.