EchoSphere

Inspiration

We didn’t come into this with personal experience caring for someone with Alzheimer’s — but we were struck by how isolating and confusing the condition can be. We asked ourselves, what if technology could bring clarity, comfort, or even a sense of connection to someone whose memories are slipping away? That idea led us to EchoSphere — a system that bridges AI, speech, and memory to support people in a genuinely human way.

What it does

EchoSphere is a voice-based memory assistant built for users with Alzheimer’s or memory loss. It:

Transcribes and summarises spoken conversations
Lets users talk to “associates” — digital versions of familiar people
Saves reminders, interactions, and settings across sessions
Uses AI to respond in a personalised, context-aware way

It’s designed to feel intuitive, familiar, and supportive — like talking to someone who remembers you even when you forget.

How we built it

Backend built with Flask
User login through Google OAuth
Data managed using SQLAlchemy with models for users, reminders, associates, settings, and conversations
Whisper handles speech-to-text transcription
Pydub is used for audio conversion and formatting
Ollama provides lightweight LLM summarisation
Created a TextToSpeechChatbot, employing Neuphonic API to drive realistic, memory-based conversations

Challenges we ran into

Audio handling (formats, sample rate conversions) took longer than expected
Hosting delayed our progress, as out attempt to find free services lead to nothing, forcing us to host locally.

Accomplishments that we're proud of

Created a full-stack application that combines voice, memory, and AI
Made the chatbot feel personal, not generic
Built and tested multiple user profiles with realistic simulated data
Designed for a use-case that feels genuinely meaningful, not just technically interesting