Pretty Body - machine learning in healthcare industry

Cloud architecture for human body 3D model predictive avatars using machine learning algorithms

Customer: a healthcare startup focused on innovation that plans to scans the human body (20000 bodies in the first phase) and identifies various body types, to then use it to create machine learning classification algorithms to classify each scan into a body type. Such classification could be used to show customers what would be their perfect body shape based on the classification they fit in and recommend best fit diet and training programs to achieve the perfect shape, showing 3D model predictive avatars on how they could look like if they respect the program in 3, 6, 9 months.

Project management: SCRUM

Overview

The customer needed a strategic development partner to design and build an innovative solution for processing and visualizing tri-dimensional scans of the human body from photogrammetric scanners. On the frontend side with chose tools and technologies to enable fast development and interactive dynamic content. On the backend side we chose tools and technologies that would ensure high availability and performance:

Technology

Frontend:

Angular, JavaScript
LESS, Bootstrap
JSON + REST

For real-time monitoring we selected:

Elastic Stack
CACTI
Zabbix

Backend:

Java, Kotlin, PostgreSQL
Spring (Cloud, Boot, Security)
JPA + Spring Data JPA
Kafka, RabbitMQ, Thymeleaf
Logstash, Prometheus

The solution

Our team has been working alongside the customer’s team from the early concept initiation phase, providing consultancy regarding the best-fit cloud architecture and the general design for their high performant microservices-based solution.

Our team designed an E2E solution that copes with an enormous amount of data to be processed for real-life scanning and analysis. Beyond being highly-performant, the solution had to be designed with a focus on usability for the three main functional design features: Administration Panel, Client Panel and Expert Panel.