Highlights
- Innovative Healthcare App with AI: designed a groundbreaking system that can centralize healthcare data and come up with one number score to reflect the overall health span of a user and help to live a longer, healthier, and happier life.
- AI Expertise at Scale: delivered our extensive expertise in Machine Learning, Computer Vision, Retrieval-Augmented Generation, and Large Language Models to develop a unique functionality in a powerful iOS app.
- 95+% Accuracy Rate of AI/ML Models: ensured the highest possible accuracy rate of the machine learning models at the core of our healthcare app, significantly contributing to cutting-edge user experience and healthcare suggestions reliability.
Client
The client is a B2C company operating in the Healthcare industry. The mission of the company is to help people take care of their health and well-being, providing feedback on the process of aging and giving actionable advice on aesthetic wellness.
This company started as a small project aimed to centralize Healthcare for greater accessibility. We, at SPD Technology, assembled an expert team with proficiency in medicine, Artificial Intelligence, and Machine Learning, as well as mathematics to help our client achieve set goals and develop this project from the very inception.
Country 
Industry 
Team Size: Product
The product is a web app and a mobile app for health and beauty monitoring, designed to serve aged 18-60 Americans, who are mostly healthy but want to monitor their health to prolong their lifespan.
The key functionalities of the product include:
- Centralized health data management, including lab analysis data
- Health indicators monitoring
- Booking appointments with doctors and clinics
- Provision of tailored healthcare recommendations based on personal data
- Predicting health trends based on the history of users’ vital indicators
- Face AI analysis
- Professional skin care shop and skin care routine recommendations.
Goals and objectives
- Help Users Understand Their Health Status Relative To Their Peers:
Develop the key product feature that defines the relative health level of a user. This involves collecting various health metrics such as results of a personal quiz, medical history, and health goals, then analyzing these metrics using Machine Learning algorithms to generate real-time personalized insights.
- Provide Timely and Actionable Health Alerts to Users:
Leverage the comprehensive approach to each user to identify specific areas for improvement for each unique health profile and send accurate notifications to users about potential malfunctions or improvements in their health indicators.
- Facilitate Access to Top-Rated Healthcare Professionals and Clinics:
Connect users with the most suitable healthcare professionals, by maintaining a database of doctors and clinics with user ratings and reviews. Users should be able to filter by specialties, insurance compatibility, and preferences, and book appointments directly through the app.
- Consolidate Multiple Health Metrics into a Single Score:
Allow the users to assess overall health with just one score. This score should be updated dynamically as new data is added, and the app should offer insights into how individual metrics influence the overall score, along with recommendations for improvement.
- Provide Users with Accurate Skin Condition Assessments Using Visual Data:
Enable the app to utilize advanced AI-based technologies including Computer Vision, Image Recognition, and Machine Learning to analyze visual data and identify issues such as acne, wrinkles, necklines etc, determine patients Fizpatrick skin score in addition to overall skin condition. Detailed information about the identified conditions should be provided.
- Guide Users to Suitable Skincare Products Based on Their Specific Needs:
Develop the functionality for an app to assess the current skin condition of a user and help to pick daily care routine products tailored specifically to their skin needs. The app should also include highly accurate educational content about skincare routines and best practices.
Challenges
- Data Aggregation: Find a way to effectively collect user data from smart devices using bioelectrical impedance to create an ML-powered score.
- Computer Vision Integration: Implement our extensive expertise in Computer Vision to develop features including facial imperfection analysis, aging tracking, and apparent age estimation.
- Offer Accurate Personalized Recommendations: deal with all challenges related to implementing a Retrieval-Augmented Generation (RAG) chatbot with a Large Language Model (LLM) to explain facial analysis results and provide personalized recommendations.
- Trend and Tendencies Detection: leverage Machine Learning algorithms and statistical methods for detecting trends in aggregated health data.
Solution
To implement this project, we assembled a team of 6 experts, including a Project Manager, 2 Machine Learning Engineers, WordPress Developer, Copywriter and Business Analyst who were later joined by iOS software developers.
During the development process, our team worked in a closed cohesion with a doctor who had a deep knowledge of Artificial Intelligence and Machine Learning, as well as a User Experience/User Interface designer on the client’s side.
To build a highly precise solution, we utilized a custom dataset labeled under the direction of medical experts for the training of our AI/ML models. To ensure the quality and reliability of data, we conducted regular data quality reviews under the guidance of medical experts involved in the project.
One of the most interesting features of this project is a facial analysis of imperfections that are closely related to aging and appearance, based on the photos of the users. We leveraged our knowledge of Computer Vision, to create a functionality for facial analysis and aging estimation and tracking as accurately, as currently possible. As a result, our custom ML algorithms allow for precise age estimation of users.
To explain the results to users, we decided to go with a Retrieval-Augmented Generation (RAG) chatbot combined with a tailored Large Language Model (LLM). In this approach, information is retrieved from a large dataset and presented as generative responses from a trained language model.
So, the RAG chatbot takes on the role of retrieving suitable information from a knowledge base before generating a coherent and contextually appropriate response using the LLM. Thanks to this, it becomes possible for the chatbot to generate detailed, easy-to-understand explanations of the facial analysis results and to provide appropriate personalized recommendations on skincare and treatments.
Our team delivered this solution in the form of an iOS application that connects user smart devices and aggregates data powered with ML models wrapped into API.
Tech Stack
-
WordPress -
React Native -
Java -
Spring Boot -
AWS -
Auth0
Our results
We delivered the full scope of the project according to the client’s expectations, developing the key functionality and all killer features of this solution:
- Market-Ready Product with 95% of AI/ML Models Accuracy Rate: we helped our client turn a business idea into a cutting-edge product, ready to get funded and become a paradigm shift in the industry.
- 90% Accuracy Rate in Detecting Facial Imperfections: achieved a reliably high accuracy rate of a computer vision algorithm, enhancing the skincare recommendation feature and introducing new ways to sustain health and vitality.
- Conducting Continuous Improvement: our Machine Learning engineers continue to work on data labeling and model optimization to maximize the performance of the custom AI/ML models.
- More Advanced Features Planned: since the product is constantly evolving, such AI-enabled features as Health Score to prolong lifespan and Health Trends to proactively identify possible issues will be developed and added soon.
Highlights Client The client is a business operating Space Needle, an observation tower in Seattle, WA, in the United...
Explore Case