DriveMate AI

Inspiration

Most mobility apps are still opened only when users need a single utility, like paying tolls, checking a route, or topping up a wallet. We wanted to change that behavior by turning VETC from a transactional app into something drivers would want to open every day.

Our core question was simple:

What if a toll-payment app could think ahead for the driver?

That idea became DriveMate AI: an intelligent driving companion that helps users decide when to leave, which route gives the best overall value, and when to refuel or top up before friction happens.

Instead of optimizing only for ETA, DriveMate is built around a more realistic daily decision model: time + toll cost + fuel/energy cost + service convenience.

What It Does

DriveMate AI transforms VETC from a toll-payment utility into a more proactive daily mobility assistant.

Key Features

Predictive Commute Planning
Forecasts commute windows and helps users decide the best time to leave before congestion builds up.
Smart Route Recommendations
Frames route choices around fastest, cheapest, and best-value options using travel time, toll cost, and fuel/energy trade-offs.
Wallet Protection and Top-Up Guidance
Predicts trip-related spending risk and prompts users to top up before toll interruptions happen.
Fuel Trend Intelligence
Detects fuel price movement and turns it into practical advice for trip timing and spending awareness.
AI Driving Assistant
Lets drivers ask natural questions like:
- What is the best route today?
- Should I top up before leaving?
- What is causing the delay on my commute?
Mobility Ecosystem Recommendations
Extends beyond routing into nearby parking, charging, and other VETC-relevant services so the assistant can recommend the next best action, not just show information.
Graceful AI Fallback
Even when live forecasting is still warming up or temporarily unavailable, the product continues working with seeded recommendations so the experience stays reliable.

How We Built It

Frontend

We built the MVP as a React + Vite web app with a mobile-first product flow. The UI includes key surfaces like the dashboard, route comparison experience, assistant page, notifications, and wallet-oriented decision prompts.

We also used Leaflet and React Leaflet to support route and map-based experiences.

Backend

We used a lightweight FastAPI microservice architecture:

a forecasting sidecar for predictive trip signals
a separate assistant service for conversational AI
frontend integration that upgrades from seeded values to live outputs when the model is ready

AI Layer

The intelligence layer combines forecasting and assistant behavior:

TimesFM powers predictive forecasting for commute windows and fuel trends
Qwen via Alibaba Cloud DashScope powers the AI chat assistant
product logic turns predictions into recommended actions such as leaving earlier, choosing a better-value route, or topping up wallet balance

Data

To make the prototype realistic, we used structured demo and synthetic data representing:

trip patterns
route options
toll-related context
fuel trend signals
parking and charging touchpoints
wallet usage behavior
recommendation states across the user journey

This let us simulate a believable daily driving assistant without needing a large live production dataset.

Challenges We Ran Into

Designing AI that felt genuinely useful instead of decorative
Turning prediction into action instead of just showing another dashboard
Combining route, wallet, fuel, and assistant flows into one simple user experience
Building a believable mobility dataset quickly for demo purposes
Keeping the app resilient while live AI services warm up or fail
Balancing hackathon speed with a product direction that could scale beyond the demo

Accomplishments That We're Proud Of

Built a clear product story for turning VETC into a daily-use mobility assistant
Integrated live forecasting into the MVP instead of stopping at static mock recommendations
Added a real conversational assistant powered by Qwen
Connected prediction with practical actions like route choice and wallet top-up guidance
Designed a product direction that expands naturally into parking, charging, and other ecosystem services
Created a prototype that is both demo-friendly and aligned with a realistic enterprise rollout path

What We Learned

AI feels most valuable when it helps users make a decision, not just consume information
Daily habit formation comes from reducing friction in recurring moments
Product framing matters as much as model capability
Reliability and fallback behavior are essential for trust in AI-assisted experiences
Even with synthetic data, a strong decision loop can make the product feel concrete and credible
The strongest hackathon demos connect technical implementation to measurable business value

What's Next for DriveMate

Improve parking and charging recommendations with more location-specific constraints
Strengthen wallet protection logic based on route risk and predicted spend
Expand post-commute recommendations such as car wash and convenience stops
Improve personalization using repeat driving behavior and prior user actions
Deepen real-time traffic and ecosystem integrations
Evolve DriveMate into a broader VETC intelligence layer for daily mobility decisions

Long-term vision:
From toll-payment utility to an intelligent mobility ecosystem built around predictive, action-oriented driving assistance.

Tech Stack

Category	Technology	Purpose
Frontend	React 19	Main user interface
Frontend Build	Vite	Fast local development and production builds
Mapping UI	Leaflet, React Leaflet	Route and map-based experiences
Backend API	FastAPI	Forecast and assistant microservices
AI Forecasting	TimesFM	Commute-window and fuel-trend prediction
AI Assistant	Qwen via Alibaba Cloud DashScope	Conversational driving assistant
Language	JavaScript, Python	Frontend and backend implementation
Data Modeling	Synthetic mobility datasets	Demo trip, route, wallet, and service behavior
Infra / Proxy	Docker, Nginx	Containerization and service routing
Cloud Direction	Alibaba Cloud ECS / Function Compute	Deployment path for scalable production rollout