Demo Application | EY Techathon 6.0 – Round 2 Submission
This project demonstrates an Agentic AI–powered Vehicle Health Intelligence Platform that shifts vehicle maintenance from a reactive model to a predictive and autonomous one.
The system continuously analyzes vehicle telematics data, predicts potential mechanical failures, automatically schedules service appointments, notifies customers, and feeds root-cause insights back to manufacturing teams for continuous quality improvement.
This repository contains a Next.js-based demo application showcasing the end-to-end intelligence flow, agent orchestration, and stakeholder dashboards.
Traditional vehicle maintenance systems operate after failures occur, resulting in:
- Unplanned breakdowns and safety risks
- High repair costs and customer dissatisfaction
- Poor service center workload utilization
- No real-time feedback loop to manufacturing teams
There is a clear need for an autonomous, intelligent system that can predict failures early and act without manual intervention.
Our solution introduces a Hybrid Agentic AI System that:
- Predicts component failures using ML models
- Coordinates autonomous decisions via AI agents
- Auto-schedules service appointments
- Communicates proactively with vehicle owners
- Generates RCA/CAPA insights for manufacturers
The demo simulates this complete workflow through a single unified web application.
Coordinates the complete lifecycle and assigns tasks to worker agents.
- Data Analysis Agent – Monitors incoming telematics data
- Diagnosis Agent – Predicts failure probability & severity
- Scheduling Agent – Auto-books optimal service slots
- Customer Agent – Sends alerts via chat/voice interface
- RCA Agent – Identifies recurring fault patterns
- Manufacturing Insights Agent – Generates CAPA-ready reports
- Security Agent (UEBA) – Monitors anomalous agent behavior
In this demo, agent behavior is simulated through controlled workflows to demonstrate orchestration logic clearly.
- Vehicle Health Score Dashboard
- Predicted Failure Alerts with confidence levels
- AI-generated service recommendations
- Simulated chat-based customer notifications
- Predictive service booking calendar
- Vehicle issue summary per appointment
- Load-aware scheduling visualization
- Component-level failure trends
- Recurring fault heatmaps
- Root Cause Analysis (RCA) insights
- Feedback loop for design improvement
- Simulate incoming vehicle telemetry
- View predicted failure on dashboard
- Observe autonomous service scheduling
- Show customer notification trigger
- Display manufacturing insight analytics
⏱️ Ideal demo duration: 2–3 minutes
| Layer | Technology |
|---|---|
| Frontend | Next.js (React) |
| UI Styling | Tailwind CSS |
| State Management | React Hooks / Context |
| Backend (Simulated) | API Routes (Next.js) |
| AI / ML (Conceptual) | scikit-learn, pandas (offline models) |
| Agent Framework (Conceptual) | LangGraph / CrewAI |
| Database (Demo) | Mock JSON / In-memory |
| Charts & Visualization | Chart.js / Recharts |
This MVP has been developed for demo purposes using the Gemini 3.1 LLM to power the AI/ML and agentic framework.
├── app/
│ ├── dashboard/
│ ├── service-center/
│ ├── manufacturing/
│ └── api/
├── components/
├── data/
│ └── sample-telematics.json
├── public/
├── README.md
└── package.json
- Node.js ≥ 18
- npm or yarn
git clone https://github.com/your-org/vehicle-health-ai-demo.git
cd vehicle-health-ai-demo
npm installnpm run devOpen:
👉 http://localhost:3000
- Telematics data is synthetic, inspired by public datasets
- Failure predictions are simulated for demo purposes
- Agent decisions are rule-driven to clearly explain logic
- Focus is on system intelligence & workflow, not raw accuracy
- Role-based dashboard segregation
- Simulated UEBA monitoring
- No real vehicle or user data used
- 30–40% reduction in unplanned downtime
- 20–25% fewer breakdown incidents
- 15–20% improvement in customer satisfaction
- Faster manufacturing defect resolution cycles
- Real-time IoT integration
- LLM-powered conversational assistant
- EV-specific predictive models
- Fleet-level analytics
- SaaS deployment for OEMs
Team Name: CurioNative
Institution: ST. Thomas Institute For Science & Technology, Trivandrum, Kerala
Built for EY Techathon 6.0 – Round 2 Submission
© 2025 CurioNative. All rights reserved.