MUNINN is a cognitive context assistant designed to support people with dementia during digital communication. The idea was inspired by a simple but difficult moment: recognizing a familiar face on a screen, knowing that person matters, but not being able to quickly recover the context needed to respond with confidence. I wanted to build something that could reduce that friction and restore a sense of comfort, identity, and connection.
While building the project, we learned a lot about balancing technical ambition with human-centered design. It was not enough to make face detection or profile retrieval work in isolation. The real challenge was making those pieces feel calm, timely, and helpful instead of distracting. We also learned how much coordination is required when combining an Electron desktop shell, a React and TypeScript frontend, an Express backend, Firebase persistence, and a recognition pipeline that depends on screen capture, face targeting, and dwell detection.
We built the project as a desktop prototype with a layered architecture. Electron handled the desktop environment and overlay behavior, React and Tailwind powered the interface, and the backend managed profile retrieval and note generation. The recognition flow watches the screen, detects faces, tracks when the user focuses on one long enough to suggest hesitation, retrieves the linked person profile, and then generates a concise “personhood” reminder with relationship cues and conversation anchors. The goal was to make the support feel ambient and unobtrusive rather than intrusive or robotic.
The hardest challenges were reliability and subtlety. Recognition is only useful if it activates at the right time, so tuning dwell behavior and keeping the detection loop stable was difficult. The overlay also had to remain visible without getting in the user’s way, which meant handling transparency, click-through behavior, and state synchronization carefully. Another challenge was shaping profile data into reminders that feel humane and specific rather than generic.
What we are most proud of is that the project is not just a technical demo. It is an attempt to use software to preserve personhood and reduce anxiety in moments of uncertainty. Building MUNINN pushed me to think beyond features and toward trust, timing, and dignity, which was the most valuable lesson in the entire process.
Log in or sign up for Devpost to join the conversation.