EchoNote

EchoNote — an AI meeting assistant that listens, summarizes, and transforms conversations into organized tasks and notes

Comment

Inspiration

Team meetings often end with scattered notes and forgotten ideas. We wanted a way to automatically capture conversations, summarize the main points, and turn them into structured insights or next steps—without anyone having to manually take notes. That’s how EchoNote was born: an AI-powered meeting companion that listens, understands, and summarizes in real time.

What it does

EchoNote records meetings from Teams, transcribes speech to text, identifies key decisions, action items, and summaries, and generates a concise note list automatically. It also organizes content by topic and speaker, and allows exporting summaries to productivity tools like Notion or Jira for easy follow-up.

How I built it

I built EchoNote using Amazon Bedrock AgentCore for agentic orchestration, Amazon Transcribe for speech-to-text, and Amazon Bedrock (Claude/Gemini models) for summarization and note-generation. The backend runs on AWS Lambda + API Gateway, while meeting data is stored securely in DynamoDB. The front end is a Next.js dashboard that lets users replay snippets, view structured summaries, and edit notes collaboratively.

Challenges I ran into

  • Fine-tuning prompts so summaries stayed concise yet accurate.
  • Handling overlapping speech and background noise in transcriptions.
  • Managing token limits for long meetings while maintaining context continuity.
  • Designing scalable AWS roles and policies for secure per-user data access.

Accomplishments that I'm proud of

  • Built a working prototype that converts a 30-minute meeting into a clean, actionable summary within seconds.
  • Successfully integrated Bedrock AgentCore primitives for task automation.
  • Created a sleek, responsive dashboard UI using Next.js and Tailwind.
  • Deployed a serverless pipeline with fully automated data flow from audio upload to summarized output.

What I learned

  • How to design agentic AI systems that can autonomously reason and call multiple AWS services.
  • Best practices for serverless data pipelines and secure AWS IAM design.
  • Techniques for speech segmentation and context preservation in LLM summarization.
  • The value of iteration—small prompt tweaks can drastically improve output quality.

What's next for EchoNote

I plan to expand EchoNote into a fully agentic meeting assistant that not only summarizes but also takes action. Upcoming features include:

  • Jira ticket automation — automatically convert detected action items into structured Jira issues with assigned owners and due dates.
  • Notion workspace integration — sync meeting summaries and key insights directly into Notion pages for easy team collaboration.
  • Real-time multi-language transcription and summarization for global teams.
  • Integration with Slack and Microsoft Teams for instant post-meeting summaries.
  • A semantic search engine powered by embeddings to query insights across past meetings.

Built With

Share this project:

Updates