See more case studies

Respire AI

TW / Dr. BreathE

An AI-powered app that analyzes sleep and detects breathing issues for better health insights

Transforming sleep health with AI

Challenge

Dr. BreathE, a health and wellness company from Taiwan, came to us with a big idea: create a mobile app to help people tackle sleep health issues like sleep apnea. They wanted to use their advanced AI model to monitor sleep, detect breathing problems, and provide actionable insights – all in a way that felt intuitive and easy for users.

Their goal was to create an app that could assess a person’s sleeping and breathing through facial image analysis and snoring detection. The idea was to offer a more convenient alternative to traditional sleep studies, enabling users to gather sleep data from home instead of undergoing long hospital stays.

The challenge? Turning these complex features into an engaging, user-friendly app that empowers people to improve their sleep health.

Solution

We partnered with Dr. BreathE to bring Respire AI to life, combining research, strategy, design, development, and quality assurance to create a game-changing health app.

We began by diving deep into the market, analyzing competitors, and running workshops to explore user needs and align features with the app’s goals. This laid the groundwork for clear user flows and features that balanced user expectations with business objectives.

In the design phase, we prioritized creating an intuitive, user-friendly experience. From wireframes to polished designs, every detail was crafted with accessibility and ease of use in mind.

For development, we chose Flutter to streamline the process, enabling us to build for Android and iOS with a single codebase – saving time and boosting efficiency. By integrating Dr. BreathE’s advanced AI model using TensorFlow Lite, we enabled the app to analyze sleep data directly on users’ devices. Smart data handling ensured smooth performance without overloading storage or memory.

We built a robust backend to support essential functionalities like account management, content delivery, and real-time sleep statistics calculation. Using Node.js and the NestJS framework, it seamlessly integrated with AWS services (Elastic Beanstalk, S3, RDS) for scalable, cost-efficient hosting. Acting as a proxy for machine learning models, it enabled accurate OSA (Obstructive Sleep Apnea) scoring through facial photo analysis.

Extensive testing with tools like BrowserStack and Firebase ensured a seamless, reliable experience. Early user feedback guided thoughtful updates, improving usability and functionality even further.

Keeping user satisfaction front and center, we conducted usability tests before the app’s launch and gathered feedback from early users through a diary study shortly after the release. The insights from these efforts guided us in refining Respire AI further through updates, ensuring it continuously meets user needs. After finishing development, we focused on ongoing maintenance to ensure the app stayed reliable and crashfree.

Results

We delivered Respire AI, a first-stage digital product that marks a major milestone for Dr. BreathE. Combining advanced AI with intuitive design, the app provides a solid foundation for addressing sleep health challenges and growing the company’s presence in Taiwan’s healthcare market.

Please note: Respire AI is a mobile app, not a Software as a Medical Device (SaMD). Together with our client, we encourage its users to consult a doctor if needed.

Dr. BreathE appreciated how we worked closely with their team, keeping communication smooth and turning their vision into a functional, user-friendly app. They valued our ability to handle both technical challenges and user needs with care and expertise.

To support our client’s next steps, we’re ensuring a smooth transition of Respire AI to a local development team in Taiwan. Dr. BreathE is now focused on growing their presence in this region and validating the app in clinical and market settings, with plans to scale globally. We’re proud to have set them up for success and look forward to future opportunities to support their growth in new markets.

Key features

Account creation

The app lets users quickly create accounts using social logins like Google Sign-In or Sign In with Apple. For those who prefer using email and password, a deep link conveniently redirects them from their email inbox back into the app, ensuring a smooth registration process.

Selfie analysis

After registering, users complete their onboarding with an AI-based breathing evaluation. They take front and profile selfies, analyzed by the AI to calculate a score. To ensure accuracy, we added error handling, text and video guides, and a feature to use volume buttons for easier profile photos.

Sleep recording

Users can detect breathing issues or disturbances with sleep recording. Sound is captured even when the app runs in the background or the device is locked. All recorded data stays on the device, and is analyzed locally. Storage is optimized by saving only summarized results in real-time.

Sleep analysis

We collaborated with our client to integrate their AI model, optimizing it for mobile platforms. The AI identifies snoring sounds, turning them into data for sleep analysis, which in turn helps users understand their breathing issues during sleep.

Results visualizations

Sleep analysis results are displayed as intuitive charts, making the data easy to understand. These visualizations highlight trends over time, with options to view weekly or monthly data. Users can explore specific days by dragging their finger across the chart.

Educational content

Users gain access to a library of educational articles and videos provided by our client, who is an expert in this field. To ensure the app stays current, we designed a system that allows the app owner to add new content without needing to release a new version of the app.

Questionnaires

The app includes questionnaires about well-being to help raise awareness of breathing issues. Completing them unlocks extra educational content, providing users with valuable insights and tools to address their health concerns.

How we did it

We started by thoroughly analyzing competitors and current market solutions to gain a deep understanding of the landscape. During competitive analysis, we looked at popular apps for tracking sleep parameters, monitoring snoring, and helping people with respiratory issues.

The research focused on the perspective of our personas – sleep apnea clinic patients. We mapped the strengths and weaknesses of the selected apps, explored how users experience recording and analyzing their sleep, and reviewed the emotions and opinions shared in user comments.

Competition analysis

Miro

Reviews

Sentiment analysis

Patient perspective

Personas

Experience testing

Applications

With the knowledge gathered during the research, we conducted a multi-day product strategy workshop with key stakeholders at our Wroclaw office. We wanted to define the product goal and strategy to achieve it, using our cross-domain expertise. This phase helps our clients avoid costly mistakes and minimize risks.

During the workshop, we created a User Story Map and a Wireflow – a visual representation of the core user flow. This helped everyone align on the key interactions and get a clear view of the product’s main flow. By the end, we were fully set to dive into development.

Mapping and collaboration

Miro

Prioritizing features

User Story Mapping

Defining product goals

Product Canvas

Focusing on user needs

Personas

Outlining user journeys

Wireflow

Highlighting key benefits

Value Pyramid

Wireframes

After the product discovery workshop, we built wireframes as a more detailed and structured product version. They provided a clear view of the app’s various flows, information architecture, and overall behavior. This step was essential for stakeholders, developers, and users to understand the product’s functionality in greater depth. It also allowed us to plan specific development tasks and test the app’s logic and usability from a practical standpoint.

Hi-fi designs

Once the wireframes were completed, we moved on to creating high-fidelity designs. We started by generating several concepts for the visual design, carefully considering cultural nuances, the latest design trends, and accessibility guidelines.

Keeping these factors in mind, we developed a clean, modern UI, which was intentionally designed in dark mode due to the app’s primary use at night. This choice helped mitigate potential sleep disruption and improved the contrast of UI elements, making the app more user-friendly during nighttime use.

After the design phase, we conducted extensive usability testing. This testing helped us gather valuable insights to guide the next steps in the product’s development, ensuring that future iterations would align with user needs and expectations.

Workshops

Whiteboard, Marker, Post-its

Flows

Miro

Wireframes, UI

Figma

Microinteractions

Principle

We chose Flutter to build the app, enabling us to create it for both Android and iOS with a single codebase. This approach sped up development and kept the process cost-effective.

To embed the client’s AI model into the app, we used TensorFlow Lite, ensuring that the AI processing runs directly on the device. To integrate the AI model seamlessly, we rewrote the client’s Python script into Dart for compatibility with the mobile platform.

For image processing, we used the CameraX package, ensuring that photos taken on various Android devices met the technical requirements specified by the AI model. This solution resolved compatibility issues with image capture across devices.

To support audio recording while the app runs in the background, we extended the record package with custom native code for both Android and iOS. The audio recordings needed to meet precise technical specifications required by the AI model.

Given the limitations of mobile device storage and RAM, we implemented a smart solution for handling extended audio recordings. Instead of recording one large file, we divided the recording into shorter segments. These segments were analyzed by the AI model immediately after being recorded. This approach minimized memory and storage usage, reduced processing time, and significantly lowered the risk of app crashes when dealing with large datasets.

We also ensured that sleep analysis data derived from the recordings wouldn’t be lost due to unexpected events like battery drain or app crashes. The app stores the analysis data locally and automatically syncs it when the app is reopened.

The selfie-capturing process required multiple iterations and close collaboration with our UX team to refine the experience. The final solution includes tools like head-placement overlays, face-outline guides, text and video instructions, and even a custom implementation for taking photos using the physical volume buttons. This feature is especially helpful for profile selfies, where users may not be looking at the screen and need a more intuitive way to capture the image.

With these tailored solutions, the app delivers a seamless and user-friendly experience, meeting both technical and user expectations.

Framework

Flutter

Architecture

BLoC

Testing

Mockito

Language

Dart

Networking

Dio

Continuous Integration

Bitrise

AI model integration

TensorFlow Lite

We built a REST API as a centralized solution to handle sleep metrics from the mobile app. Our goal was to enable real-time aggregated statistics that consider users’ time zones, ensuring they can access detailed insights anytime, anywhere.

We implemented the solution using Node.js and the NestJS framework, hosting it on AWS with cost-efficient services like Elastic Beanstalk, S3, and RDS. To ensure precise time-based calculations, we developed a custom DateTime handling module that efficiently manages time zones. This architecture enabled fast feature development, seamless scaling, and accurate data processing.

To guarantee high-quality performance and reliability, we conducted thorough end-to-end testing with Jest and Testcontainers. The API we designed delivers scalable and robust functionality, providing users with valuable insights to improve their sleep health.

What’s more, the backend acts as a proxy for an external machine learning model that analyzes facial photos to generate OSA (Obstructive Sleep Apnea) scores, helping assess sleep apnea risks. We streamlined this process using AWS S3 to ensure smooth and efficient data management throughout the system.

API type

REST

Language

Typescript

Database

PostgreSQL

Framework

NestJS

Database framework

Drizzle

File storage

AWS S3

End-to-end tests

jest + Testcontainers

To ensure top-notch quality, we didn’t just rely on manual testing – we also used advanced tools to streamline the process.

  • We leveraged Jira and Xray to manage test cases and track defects efficiently, along with Testpad to design and organize tests.
  • For device compatibility, BrowserStack gave us access to a wide range of real devices and emulators, enabling us to test the application across multiple platforms and configurations.
  • When it came to backend testing, we used Postman, Insomnia, and Swagger. These tools helped us quickly validate backend changes, catch issues early, and respond effectively.
  • We also integrated Firebase to track how users interact with different areas of the app. This gave us insights into which features are the most popular and which might need more attention, helping us prioritize future development.

In short, these tools helped us to deliver a high-quality app, ensuring a smooth user experience and driving engagement.

Test management

JIRA, Xray, testpad

Devices farm

Browserstack

API testing

Postman, Insomnia, Swagger

Analytics

Firebase

Before releasing the app, we conducted usability tests to see how users navigate features, understand data collection (e.g., breathing sounds and photos), and interpret instructions for health indicators like snoring or apnea indexes. Users appreciated the straightforward issue-selection flow, ease of sleep recording, and intuitive snoring results.

However, improvements were needed in simplifying medical terms, clarifying data requirements, providing step-by-step photo-taking guidance, and explaining sleep recording nuances like phone placement. Based on these insights, we identified key adjustments to enhance usability and ensure the app meets user expectations. These updates were prioritized for future iterations.

Collaboration

Google Meet

Documenting findings

Google Docs

Design updates

Figma

After releasing the app, we conducted a diary study with early users in traditional Chinese to understand how it fit into their routines and navigation behavior. Feedback highlighted a straightforward registration process and user-friendly onboarding. However, challenges emerged, such as difficulties with side profile photo-taking, unclear neck size graphs, and a lack of clarity in the daily sleep breathing pattern summary. Users also needed better explanations about the purpose of collected data like neck size and photos.

To address these issues, we introduced realistic neck size visuals, improved the photo-taking process with sample images and helpful tips, simplified the summary screen with bullet points, added an educational screen linking data to sleep risk, and refined UX writing across the app. These updates enhanced usability and user satisfaction for everyday use.

Collecting feedback

Typeform

Analyzing the data

Google Spreadsheets

Design updates

Figma

How it works

Testimonial

Working with Droids On Roids over the past six months has been a smooth and collaborative experience. Together, we’ve achieved key milestones, including integrating AI image recognition, audio analysis, and medical surveys, all while focusing on creating a secure and user-friendly app. The agile approach, with regular updates and clear communication, helped me stay closely connected to the progress. I’ve truly enjoyed discussing ideas and solving challenges with such an enthusiastic and professional team.

柏傑
Product Manager & AI Engineer, Dr. BreathE

Looking to create a health app?

Let our expertise guide you every step of the way!

Let's talk or just write an email
Not ready for an estimate? Write to us and book a Product Design Workshop to clarify your vision and prepare the crucial documentation.

Other case studies

Hoppy – Product Distribution App

Reward-driven product distribution app. Made for Carlsberg Polska

Nice To Fit You – mCommerce Diet Catering App

mCommerce diet catering app

Portfolio

epyGuard™ | IoT App

Connectivity app tracking heat exhaustion levels in non-desk workers

Firm Prospects

Large attorney-only community for in-depth law firm research, job listings and career updates.

GYDER

Easy-to-use event platform that helps visitor to make the most of their time at an event.

See more apps
See more case studies

FAQ on sleep tracking app development

A sleep tracking app is a mobile application designed to monitor and analyze users’ sleep patterns. These apps use features like sound recording, motion detection, or wearable integration to provide insights into your sleep quality, duration, and disturbances, helping you improve your overall sleep health.

A sleep app helps by tracking sleep patterns, identifying disturbances like snoring or apnea, and providing actionable recommendations. Advanced apps, like Respire AI, use facial image analysis and sound recording to give users deeper insights into their sleep health.

Modern sleep tracker apps leverage AI and machine learning to provide highly accurate data. While they may not replace medical-grade sleep studies, apps like Respire AI offer a convenient alternative for monitoring sleep health at home.

Yes, many sleep recorder apps use advanced algorithms to analyze nighttime sounds and identify patterns that indicate snoring or sleep apnea. Respire AI, for example, provides Obstructive Sleep Apnea (OSA) scoring to assess sleep health risks.

A sleep tracking app offers the convenience of monitoring sleep from the comfort of your home. While traditional sleep studies are more comprehensive, apps like Respire AI provide a user-friendly and cost-effective way to track sleep metrics daily.

Health app development requires a combination of research, strategy, and user-centric design. For a sleep monitor app, this includes leveraging AI, ensuring data accuracy, and creating an intuitive user experience. Carefully analyzing and choosing the right development partner significantly increases your chances of success in the market.

Most sleep tracker apps store and analyze data locally or securely in the cloud. Apps like Respire AI focus on privacy by performing most analyses directly on the device and implementing robust data encryption.

Investing in a sleep app addresses the growing demand for convenient health solutions. A well-designed sleep app not only helps users improve their sleep but also taps into the rapidly expanding health tech market.

If you’re considering health app development, our case study on Respire AI showcases our expertise in building innovative sleep tracking applications. Let’s talk about your project!

Medical app development often requires adherence to strict regulations and standards, such as HIPAA or GDPR, to ensure data privacy and security. It also demands a focus on integrating features like remote monitoring, electronic health records (EHR), and AI-powered diagnostics.

Healthcare mobile application development comes with challenges like maintaining data security, meeting regulatory compliance, and ensuring seamless integration with existing medical systems. Overcoming these challenges requires technical expertise and industry knowledge.

Medical app development enhances patient care by offering real-time health tracking, remote consultations, medication reminders, and access to medical records. These features improve patient engagement and health outcomes.

Partnering with an experienced team for health app development ensures the app is well-researched, user-focused, and compliant with industry regulations. This approach increases the app’s chances of success in a competitive market.