Inspiration
As new parents, keeping track of everything discussed during a pediatric visit can be overwhelming. Medical terms, instructions, and next steps are often hard to remember, especially when dealing with a fussy baby. We wanted to create a tool that empowers parents by capturing and summarizing these conversations clearly — and even goes a step further by offering personalized, age-specific baby care insights. That's how CuddleScribe was born.
What it does
CuddleScribe is an AI-powered assistant for pediatric visits. It allows parents to:
- 🎙️ Record or upload conversations with their pediatrician
- ✍️ Get an instant, AI-generated summary of the discussion
- 📌 Receive personalized baby care tips and milestones based on age
- 🕒 Get automated email reminders for future doctor sessions It's like having a smart medical scribe and parenting guide, all in one.
How we built it
CuddleScribe is built entirely on AWS serverless services, making it scalable, lightweight, and cost-effective.
- Frontend: Built with React for a clean, user-friendly experience
- Backend: AWS Lambda functions handle transcription, AI summary generation, and email reminders
- Speech-to-text: Amazon Transcribe
- LLM Summarization & Insights: Amazon Bedrock (Nova Pro)
- Storage: Amazon S3 for audio files
- Database: Amazon DynamoDB
- Email Notifications: Amazon SES,
- Scheduling: EventBridge scheduler
Challenges we ran into
- Prompt tuning to generate summaries that are concise yet informative for parents
- Navigating SES sandbox limitations during development
- Designing a user flow that feels simple despite integrating multiple AWS services
- Provide real-time updates to end-user after async file upload & processing
Accomplishments that we're proud of
- Delivering a working end-to-end serverless product in a short time
- Building a smooth voice-to-summary pipeline
- Integrating meaningful baby development info and reminders into the experience
- Creating a user-friendly UI that parents can trust and use with ease
What we learned
- How to combine serverless AWS tools to build a fast, scalable, AI-powered solution
- Prompt engineering for domain-specific summaries (in this case, healthcare)
- Optimizing Lambda workflows for real-time streaming and batch processing
- Importance of designing for accessibility and trust in medical applications
What's next for CuddleScribe
- Build a mobile app for parents to record visits on the go
- Add calendar integrations and smart reminders from summaries
- Use NLP to detect possible follow-up actions or questions to ask
- Explore end-to-end encryption & HIPAA compliance for production use
- Partner with pediatricians to test real-world feedback and pilot the tool in small clinics
Built With
- ai
- amazon-dynamodb
- amazon-web-services
- appsync
- bedrock
- cdk
- graphql
- javascript
- lambda
- node.js
- react
- s3
- serverless
- transcribe