Inspiration
Over 70% of adults aged 40 and older experience difficulty sleeping, which is strongly linked to fatigue, stress, and reduced overall well-being. Many sleep apps and trackers focus solely on metrics—heart rate, latency, or breathing—but for people who already feel anxious about their sleep, seeing all that data can actually worsen their current condition. These experiences prompted the launch of DormiCare, an app that provides personalized advice, simple routines, and gentle support to help people of all ages, particularly older adults, improve sleep, reduce stress, and feel more in control of their daily wellness.
What it does
DormiCare provides personalized advice, simple routines, and gentle support to help people of all ages, particularly older adults, improve sleep, reduce stress, and feel more in control of their daily wellness.
How we built it
I coded the FastAPI backend in python which is mentioned on my repository in github. The frontend was vibe-coded using react iOS. My cofounder added a customizable UI/UX design to the frontend using react and rust, and I coded the asyncrhonous processing of the backend which would lead into the PostGRE SQL database storing the user queries like sleep logging or their responses to our initial survey in the app after signing in through their apple account. Multi-LLM API fallbacks were implemented using the Grok and OpenAI APIs in order to ensure that the user recieves a response from either LLMs, and pydantic schemas were implemented on the queries to ensure that the user's queries were valid numbers or explanations according to the survey questions or responses accepted.
Challenges we ran into
Users were mentioning in the feedback surveys we collected that sleep log updates weren't being updated simultaneously on both Apple devices they were using at the same time. This led us to implementing the last-write-wins policies in our code for the backend in order to ensure that the server payload and the external-version payloud queries were registering at the same time to prevent discrepencies in user data.
Accomplishments that we're proud of
Our app reached over 150+ users in its early release period and 1.6k impressions over the course of a few months since we developed the app this year. Senior caregivers at the veterans memorial home in addition to adult caretakers reported higher focus levels and improved memory retention in veterans through our feedback forms. We also wrote pytest suites for our app in order to identify and test our app in deployed environments for testing for bugs.
What we learned
The purpose of this app was inspired by Elena, a 67-year-old veteran’s wife at the Menlo Park Veterans Memorial Home, who was frustrated and uncomfortable in a cast that she was required to wear, with no one to listen. That experience made us realize that solving real-world problems isn’t just about highly-efficient medical appliances—it’s about understanding the person using it. Consulting with a renowned sleep psychologist also helped us see how sleep anxiety affects people and why comfort and support matter as much as guidance.
What's next for DormiCare
We aim to expand our parnerships for our company DormiCare through consulting with other sleep psychologists, and working with other medical professionals who are more experienced with working with patients suffering from orthosomnia. After consulting with Dr. Shelby Harris about discussing current sleep trends in her clinic and the focus that should be directed toward sleep-anxiety triggering quantitative metrics, we believe expanding our brain games in the app and adding additional LLMs or integrating docker to containerize our app.
Built With
- docker
- faiss
- fastapi
- git
- github
- nltk
- numpy
- pandas
- pydantic
- pytest
- pytorch
- rest-apis
- scikit-learn
- sqlalchemy
- swagger/openai-api
