SkimSage — Autonomous AI Study Coach

AWS Bedrock-powered autonomous AI study coach that transforms lecture notes into personalized learning plans and adaptive quizzes.

Comment

Inspiration

We wanted to solve a real student problem — spending hours summarizing lecture notes and making quizzes manually. With AI advancing so rapidly, we thought:

“What if studying could be autonomous — an AI study coach that turns notes into personalized learning experiences?”

That idea became SkimSage — built for the AWS AI Agent Global Hackathon, powered by Claude 3.5 Sonnet through AWS Bedrock.

What it does

SkimSage is an autonomous AI study coach. It:

  • Ingests lecture notes (PDF, TXT, or S3)
  • Extracts and summarizes key concepts
  • Generates interactive quizzes and learning plans
  • Adapts difficulty dynamically
  • Stores structured JSON results in Amazon S3 for analytics

Everything runs autonomously — from parsing → reasoning → planning → quizzing → storing — without human prompt engineering.

How we built it

Frontend: Streamlit for the interactive UI Backend: AWS Lambda + API Gateway for the inference endpoints AI Reasoning: Claude 3.5 Sonnet via AWS Bedrock (us-west-2) Storage: Amazon S3 for lecture files and results Agent Core: Python-based orchestration pipeline using Bedrock SDK Deployment: AWS SAM for reproducible infrastructure

The AI pipeline invokes Claude to analyze text, generate structured JSON, then feeds results into the UI for dynamic quiz and learning plan rendering.

Challenges we ran into

  • Configuring Bedrock permissions and IAM roles for API invocation.
  • Parsing large PDFs while keeping prompt size within Bedrock limits.
  • Merging JSON-based model outputs safely to avoid malformed responses.
  • Debugging Lambda → Bedrock → Streamlit communication under time pressure.

Accomplishments that we're proud of

  • Fully functional autonomous AI workflow — no manual prompts.
  • Real-time Claude 3.5 Sonnet reasoning through AWS Bedrock.
  • Seamless fallback mechanism for offline/local analysis.
  • Smooth Streamlit UI integration with live scoring and analytics.
  • Successfully deployed and demoed on https://skimsage.streamlit.app/

What we learned

  • How to use AWS Bedrock’s AgentCore and Claude for JSON-driven reasoning.
  • Practical design of autonomous agents that plan and adapt dynamically.
  • How to deploy serverless AI architectures with minimal latency.
  • Importance of graceful error handling and fallbacks in production AI systems.

What’s next for SkimSage — Autonomous AI Study Coach

  • 🗣️ Voice-based interactions (spoken quizzes)
  • 📈 Dashboard analytics for progress tracking
  • 🧩 Integration with Learning Management Systems (LMS)
  • 🧮 Spaced repetition algorithms for adaptive retention
  • 🎓 AI tutor persona using Bedrock multimodal models
Share this project:

Updates

Submission history