MIT xPRO Capstone
Revolutionizing Vehicle Ownership: Designing an AI-Powered Car Management Platform
A product concept focused on reducing the friction of fragmented vehicle ownership through one integrated mobile experience.
Problem
Vehicle ownership was fragmented across disconnected tools and dealership dependencies.
Owners had to manage multiple apps, rely on dealerships for key moments, and had limited ways to anticipate maintenance needs or repair expenses before a problem became urgent.
The product opportunity was not just adding AI features. It was creating a single place to make ownership feel less reactive and more manageable.
Outcome Focus
Lead with confidence, convenience, and fewer ownership surprises.
- Reduce dependency on fragmented systems.
- Help drivers anticipate maintenance earlier.
- Give multi-vehicle households one consolidated view.
- Extend relevance to EV ownership with charging support.
Solution
An integrated mobile application combining health signals, diagnostics, and ownership guidance.
Real-time health monitoring
Machine learning reviews vehicle signals to surface emerging issues sooner.
Predictive maintenance alerts
Sensor data analysis helps estimate when service may be needed before breakdowns occur.
Driving habit insights
The model works like a recommendation engine that learns from each driver's habits and highlights fuel efficiency opportunities.
Diagnostics + EV support
Pre-trained audio and image models assist diagnostics, while charging station locators support EV workflows.
Double Diamond
Two phases of divergent and convergent thinking structured the concept.
Diamond One
Discover to Define
- Identified stakeholder needs and defined KPIs.
- Converged on fragmentation as the core challenge.
Diamond Two
Develop to Deliver
- Proposed an OEM partnership strategy.
- Shaped a three-stage rollout: data collection, MVP development, and launch with iteration.
Concept Flow
A simple input-to-output model kept the AI story understandable.
Inputs
Sensor data, driving patterns, vehicle history, audio, and images.
Processing
Pre-trained models and machine learning pattern detection evaluate maintenance and diagnostic signals.
Outputs
Alerts, recommendations, driving insights, and consolidated ownership guidance.
AI Product Decisions
Product management tradeoffs centered on trust, responsiveness, and failure handling.
Success metrics
Track alert usefulness, engagement with maintenance guidance, and repeat use of diagnostics and ownership features.
Accuracy vs latency
Recommendations need to arrive quickly enough to be useful without lowering confidence in the signal.
Failure modes
False alarms, missed maintenance signals, and weak diagnostics need clear explanations and fallback guidance.
Cold start
Early usage needs enough data collection to make insights useful before behavior patterns become rich.
Six-Month Plan
From concept to evidence.
- Build machine learning models.
- Integrate the models into the app backend.
- Establish feedback collection loops.
- Explore anomaly detection for unusual vehicle behavior.
- Explore reinforcement learning for increasingly tailored recommendations.