DDnD – Don’t Drink and Drive AI Companion

Real-time AI safety assistant that detects if you're unfit to drive using Apple Watch health data built with AWS Bedrock, DynamoDB, and HealthKit. Alerts you before risky decisions.

Comment

🚗 DDnD – Don’t Drink and Drive AI Companion

💡 Inspiration

Every year, thousands of lives are lost to preventable DUI-related accidents. We wanted to create a real-time, wearable safety assistant that empowers users with biometric insights before they drive. With Apple Watch always on the wrist and AWS Bedrock providing powerful AI, we built a system to stop risky decisions before they happen.

🚀 What It Does

  • Collects live biometrics using HealthKit on Apple Watch
  • Sends data every 10 seconds to AWS via REST API
  • Uses Amazon Titan (via Bedrock) to classify driving fitness as:
    • ✅ Normal
    • ⚠️ Moderate
    • 🚫 High Risk
  • Displays risk on Watch UI
  • Pulses haptics for high risk
  • Triggers SNS alerts if necessary

🔧 How We Built It

  • Frontend (watchOS):

    • SwiftUI + HealthKit + HKLiveWorkoutBuilder
    • WKExtendedRuntimeSession for background updates

  • Backend (AWS):

    • API Gateway + Lambda (Python)
    • Amazon Bedrock (Titan)
    • DynamoDB to store risk + vitals
    • CloudWatch logging
    • SNS alert for critical risk

⚠️ Challenges We Ran Into

  • Live HR streaming was tricky — required workout sessions
  • Decimal handling in DynamoDB (float not supported natively)
  • Prompt tuning Titan to return only normal, moderate, or high
  • Lambda runtime permissions + Bedrock model access

🏆 Accomplishments We’re Proud Of

  • End-to-end pipeline works in under 10s
  • Watch app haptics + animation + real alerts ✅
  • Titan prompt response consistent and explainable
  • Used real biometric input (not just mock data!)
  • Built this across both frontend & backend in under a week

📚 What We Learned

  • How to prompt Amazon Titan on Bedrock for deterministic output
  • Real-time watchOS development and biometric streaming
  • AWS security/role setup for Bedrock and DynamoDB
  • Visualizing biometric trends in a meaningful way
  • Building health tech with AI that can save lives

🔮 What’s Next

  • iOS dashboard with risk history + trend insights
  • Push alerts to caregivers or emergency contacts
  • Trip logging and smart recommendations
  • Integrate with cars (e.g., block ignition if risk is high)
  • Expand dataset and refine Titan prompt using real-world inputs

🛠 Built With

  • SwiftUI / HealthKit
  • AWS Lambda (Python 3.12)
  • Amazon Bedrock (Titan Text Express)
  • API Gateway, DynamoDB, SNS, CloudWatch
  • boto3, decimal.Decimal

🎥 Demo Video

📺 Watch the Demo on YouTube

🔗 Live API Test

GET /getRisk?short_user_id=1234567890
Host: https://x3lurwrtk3.execute-api.us-east-1.amazonaws.com/prod

Built With

Share this project:

Updates