Elevator Pitch

Anchor: “Bringing clarity to confused minds, one real-time moment at a time.”

Anchor restores real-time context for people with dementia by turning life events — like who’s nearby, where they are, and what’s happening — into calm, text-based explanations. Caregivers get a live dashboard to monitor, add memories, and help their loved ones feel grounded. Powered by Confluent Cloud for real-time event streaming and Google Gemini for AI context, Anchor transforms confusion into clarity instantly.

About the Project

Inspiration

People with dementia often experience sudden context collapse — forgetting where they are, who’s around, or what’s happening. This triggers anxiety, repeated questions, and stress for caregivers. We wanted to create a tool that restores clarity in real time, helping both the person and their caregiver feel grounded and supported.

What it does

Anchor is a real-time context continuity engine that:

Streams real-world events (presence, calls, location, time-of-day) via Confluent Cloud.
Synthesizes context with Google Gemini into short, clear explanations.
Provides dashboards for users and caregivers to visualize context, manage people, and add memories.
Preserves memories permanently, allowing replay for continuity.
Is accessibility-first, with large text, high contrast, keyboard navigation, and screen-reader support.

How we built it

Event Streaming: Confluent Cloud Kafka topics handle all real-time events.
Backend Processing: Python + Flask backend consumes events and manages the current context state.
AI Context Generation: Google Gemini generates compassionate, understandable text-based explanations of the user’s current reality.
Frontend Dashboards: Web dashboards show context and memory history to both users and caregivers.
Persistent Memory Storage: Confluent + SQLite ensures events and caregiver-added memories are permanently stored.

Challenges we ran into

Simulating real-time events for the demo without actual IoT devices.
Generating context explanations that are short, clear, and reassuring for someone with dementia.
Integrating Confluent + Gemini + frontend dashboards within a short hackathon timeframe.
Designing accessible dashboards that are easy to navigate for users with cognitive impairments.

Accomplishments that we're proud of

Fully functional real-time event pipeline via Confluent Cloud for MVP events.
AI-generated context explanations that reduce user confusion.
Accessible dashboards for both users and caregivers.
Scalable architecture ready for IoT sensors, voice integration, and multi-language support.

What we learned

How to stream, consume, and process real-time events with Confluent Cloud.
How to leverage Google Gemini for generating human-readable context from structured event data.
The importance of accessibility and UX when building for cognitively impaired users.
How to build an MVP that addresses real-world pain points effectively under time constraints.

What's next for Anchor

Voice integration using ElevenLabs to allow hearing context instead of reading.
Mobile app support for caregivers and users on-the-go.
IoT integration for real-world presence and location detection.
Multi-language support for global accessibility.
Caregiver alerts for unusual patterns or confusion events.

Built With / Tools

Languages & Frameworks: Python, Flask, HTML, CSS, JavaScript
Cloud Services: Google Cloud Vertex AI / Gemini, Confluent Cloud (future: Cloud Run deployment)
Database: SQLite (for memory storage)
Other Tools: Kafka, WebSockets for real-time updates