EverCare

Inspiration

Dementia widely impacts families; every team member has a family member living with some form of dementia. We all wished someone could be there for them when we couldn't be, which fuelled our inspiration to build EverCare.

What it does

EverCare is an iOS app that supports people living with dementia by detecting their confusions and guiding them through their routines. It detects user confusions by using the camera and microphone to take in their facial expressions and speech, then analyzing them using Presage’s technology. It then guides users by using Cohere to generate supportive responses and uses ElevenLabs to vocalize them gently. The app also connects with users on a personal level by remembering their conversations, using MongoDB to store them for future reference. Finally, the app supports an emergency help button, prompting immediate assistance in case of a medical emergency.

How we built it

After realizing our goal to help dementia patients, we brainstormed the most helpful features that a project could provide, and discussed until we reached a consensus on the details of our project. We then delegated roles to each team member; one primarily handled frontend while the other three handled backend and integration. We also bent these roles to adapt to our situations as needed; backend developers handled different APIs over time, and one team member pivoted to working on the project’s Devpost pitch via mobile after their laptop completely ceased to function.

Challenges we ran into

Gemini filters responses that are potentially offensive, but this function heavily backfired on us since it would refuse to respond under some medical situations. In particular, user input that included sensitive mental health topics would typically receive no response, which was a risky malfunction. As a result, we eventually switched to different AI services that filtered less of their content.

Accomplishments that we're proud of

This hackathon was marked with growth for our team. We learned how to use many technologies on the fly; most of our tech stack in this hackathon were completely brand new to us. For example, Swift—the backbone of our frontend—was foreign to the entire team until the hackathon started. We’re also happy to have navigated around some technical difficulties. ElevenLabs only worked for two of our laptops and one team member’s laptop rendered itself useless by constantly shutting down, but we still managed to finish our tasks as intended.

What we learned

To power EverCare, our team members had to learn how to integrate Presage, Cohere, and ElevenLabs, MongoDB, FastAPI, and more technologies. Of course, we also had to solve countless bugs—some of which AI couldn't solve either. One of our biggest lessons was to edit broadly—we solved many problems with our AI API by switching our integration from Gemini…to Ollama…and finally, to Cohere.

What's next for EverCare

The next phases of EverCare focuses on improving reliability, personalization, and real-world impact. EverCare will expand its understanding of user routines by learning daily patterns over time, allowing it to anticipate confusion before it escalates. This includes proactively reminding users where they are, what time it is, and what they were about to do, in a calm and familiar way. EverCare also aims to collaborate with healthcare professionals and dementia care organizations to validate its effectiveness, refine its guidance strategies, and ensure the app aligns with real clinical and caregiving needs. Together, these steps move EverCare from a helpful assistant to a trusted daily companion for people living with dementia.