Somnia

home

Inspiration

Democratizing Wellness. Sleep tracking is an important part of our health, and many premium services like smartwatches and the Calm app addresses it. But we were concerned by the prohibitive costs of their premium subscriptions. We believe that mental health support and sleep hygiene should not be a luxury restricted by a paywall. Somnia was created to bridge this gap, providing high-quality wellness tools and data analysis at zero cost to the user, ensuring that financial status is never a barrier to a better night's rest.

What it does

Somnia serves as a comprehensive wellness companion designed to mitigate stress and optimize sleep hygiene through data-driven insights. Key features include:

Somnia's core analysis tool is driven by SomniQ. SomniQ leverages state-of-the-art neural intelligence trained on an Oxford-funded clinical research dataset to deliver next-generation sleep diagnostics. Through extensive fine-tuning and validation, the model learns deep patterns in sleep behavior that traditional approaches often miss, enabling scalable, high-confidence insights for both individuals and care providers.
Privacy-First Access: Full support for anonymous authentication, allowing users to prioritize their health without compromising their personal identity.
Detailed Biological Profiling: Users can establish a baseline by inputting essential metrics including age, gender, height, and weight.
Predictive Pattern Analysis: The core of the app analyzes the complex interplay between biological profiles and lifestyle data, specifically focusing on:
Sleep and exercise duration.
Daily step counts and physical activity levels.
Resting heart rate and subjective stress levels.

How we built it

We engineered Somnia using a modern, scalable architecture designed for high performance and universal accessibility:

Frontend: Developed with React Native, ensuring a fluid, native user experience across iOS, Android, and Web platforms from a single, unified codebase.
Backend: Leveraged Supabase to handle real-time database management and secure user authentication seamlessly. Used Railway to deploy our nodejs backend, with our python micro service for giving users insights about there sleep. Deployed our app to Testflight to demo on iOS and deployed to firebase for hosting.
Machine Learning: Built a lightweight ensemble model of Tree-based regressors and Deep Learning models to predict and personalize sleep-related insights from user signals (e.g., sleep duration/quality factors). The model is trained on a dataset compiled from Oxford funded scientific studies. The inference runs in our Python microservice, exposed via a secure API to the Node.js backend, enabling fast, scalable, and updatable ML-driven recommendations without changing the client.

Challenges we ran into

Overcoming Collaboration Friction. During the development process, we faced significant hurdles with version control and persistent merge conflicts. To resolve this, we pivoted our workflow to utilize Visual Studio Live Share. This allowed our team to engage in synchronous, real-time development on a shared environment, effectively eliminating code silos and transforming our bottleneck into a highly efficient collaborative process.

Accomplishments that we're proud of

Refined UI/UX: We successfully designed a minimalist, intuitive interface that achieves aesthetic and functional parity with premium, paid competitors.
Custom Intelligence: We are particularly proud of developing and integrating our own trained model to analyze sleep patterns using specific datasets, providing users with sophisticated insights into their wellness.

What we learned

This project provided a deep dive into the practical application of machine learning, specifically in how to train and implement a model for biological data. Furthermore, we gained invaluable experience in agile collaboration, discovering how real-time sharing tools can drastically improve team velocity and code integrity.

What's next for Somnia

Our roadmap for the future of Somnia focuses on expanding its wellness ecosystem through:

Multimodal Content: Integrating a library of relaxation soundscapes and guided wellness videos.
Hardware Ecosystem Integration: Expanding our data reach by syncing with smartwatches and fitness-tracking applications to automate health monitoring.