Inspiration
We’ve all been in meetings, study groups, or therapy sessions where important points were said… and later forgotten. Existing note-taking tools often miss context or lose the “thread” of conversations. We wanted something smarter—an AI that actually remembers, like a second brain.
What it does
EchoTwin is a personal AI memory assistant that listens to conversations and builds a personalized memory bank for you. It can: Summarize conversations into key takeaways. Generate follow-up tasks automatically. Remind you later in context (e.g., “Last week, your team discussed X…”). It’s like having your own AI twin who never forgets.
How we built it
Frontend: React for a clean, interactive UI. Backend: Flask for API integration. Speech Processing: Whisper for speech-to-text. Intelligence: GPT for summaries, contextual insights, and reminders. Memory Storage: Vector DB (Pinecone/FAISS) to store and recall conversations.
Challenges we ran into
Getting Whisper transcriptions accurate in noisy environments. Designing memory retrieval so responses felt contextual, not generic. Balancing latency between transcription, summarization, and storage. Making the system scalable for long conversations.
Accomplishments that we're proud of
Built an end-to-end prototype that can listen, remember, and recall. Designed a memory system that feels truly personalized. Integrated multiple AI models into a single seamless workflow. Created a UI that’s simple enough for meetings and personal journaling.
What we learned
How to combine LLMs with vector databases for contextual recall. The importance of chunking and embeddings in long conversation handling. That memory in AI isn’t just storage—it’s about usable context. Realized how valuable this could be for students, professionals, and therapy.
What's next for EchoTwin
Add multimodal support (images, slides, whiteboard snapshots). Build mobile & browser extensions for real-time meeting use. Enhance long-term memory with reinforcement learning. Explore integrations with Slack, Zoom, Notion, and calendar apps.
Built With
- amazon-web-services
- contextual-recall-vector-db-(pinecone-/-faiss)-?-memory-storage-&-retrieval-cloud-deployment-?-for-scalability-(e.g.
- react-?-front-end-ui-flask-?-backend-api-openai-whisper-?-speech-to-text-transcription-gpt-?-summarization
- render
- task-generation
Log in or sign up for Devpost to join the conversation.