Request a call
  • Hidden

Business Overview

Our client is a health-tech startup aimed to disrupt the industry with unique fertilization treatment. The company invented a methodology to identify the best embryos with the highest implantation potential using AI-powered image recognition algorithms. In combination with computerized workflows and data-driven decision support, this approach completely transforms the fertilization process for all stakeholders – clinicians and patients.

The company was looking for a technology partner to turn their method and prototype into a market-ready product and selected NIX as a proven expert in digital platform engineering, artificial intelligence (AI), and machine learning (ML).

In just 3 months NIX successfully designed and delivered an MVP of cloud-based SaaS applications for fertilization clinics. The scope for NIX included 3 key deliverables:

  • Production-ready AI engine to classify embryo images, compute predictions and assist the decision process 
  • Scalable, multi-tenant, service-oriented backend, including data orchestration and data processing
  • Modern and intuitive user interface for clinical teams to help reduce medical errors, standardize the fertilization process, and provide tangible, data-driven insights via reports and dashboards

Solution

Working side by side with the client’s subject matter experts and researchers, we enhanced the efficiency and accuracy of AI algorithms, identified image classification models, and deployed the models on the TensorFlow Serving framework. 

This system is perfectly tuned for serving trained models, scalable API, and auto-allocation of hardware resources. The models were trained on a dataset of time-lapse videos of human embryos developing.

author photo

Vitaliy

Lead Data Scientist

We leveraged a proven image classification approach that we perfected earlier on similar tasks—in turn, the client’s team brought on a table 250 GB of real-life data sourced from the leading clinics from all over the world. This is an ideal combination to achieve required accuracy and performance.

The solution features infrastructure-agnostic and cloud-ready system design based on microservices, containerization, and automated deployment. The application frontend utilizes React JS, which is well-suited to deliver a perfect user experience. 

The backend relies on Golang technology, which has proven itself as an excellent choice for intense parallel computation. 

author photo

Alex

Solution Architect

Why Golang? It’s perfectly optimized for creating high-loaded applications and a great fit for microservice architectures. The SaaS model is the best option to deliver high-tech tools for clinics and labs and enable clinicians to meet the growing demand for fertilization treatments worldwide. We expect that the number of organizations using the solution will grow substantially, so the load on the system will also increase. This is where Golang and cloud-ready design helps us maintain the robust performance of computations and ensure scalability of the system.

The client carried out a comprehensive user survey. Based on its results, the NIX team identified proto-personas and formed customer journey maps. This allowed us to expand functionality and make it accessible and efficient for non-technical users. 

As a result, the application provides the following functions: 

  • Support for different user roles: embryologists, reproductive endocrinologists (REI), and lab managers

  • A custom media player to visualize the progress of embryo development with information about the status, parameters, and quality of the embryos

  • Statistics dashboards with main indicators, historical data, and insights on fertilization and embryo development

  • Treatment cycle statuses—“Lab Report Approved/Update Needed/Created”, “Embryo Grading Ready”, “Embryos Growing”, “Treatment Completed”

  • Notifications to signal users about when a treatment report is ready, changed, approved, or rejected for corrections

Outcome

The client received a market-ready MVP of the product with sophisticated AI features and an intuitive UI that helps to reduce medical errors and allows for individualized, data-driven decisions throughout the fertilization process. 

After being trained on thousands of digital embryo images and videos, the product can reliably predict embryo quality with minimal or no human involvement and minimize the number of embryos for implantation (1-2 vs 3-4 in the standard fertilization process). This reduces the risk of multiple pregnancies and the number of fertilization attempts and offers precise quantitative and qualitative metrics through the process, therefore providing higher transparency, less stress, and less impact for patients compared to the traumatic, stressful, and expensive traditional fertilization techniques.

Working side by side with the client’s team, NIX built the platform on schedule, and the client’s team raised the next round of investments.

600x405 (14)

Team:

10 experts (Project Manager, Business Analyst, 3 Full-stack (Go + React) JS Developers, 2 Data Scientists, QA Engineer, Graphic Designer, UI/UX Designer)

Tech Stack:

Web: Golang, Echo framework, PostgreSQL, Redis, Swagger, ReactJS, Chart.js, Jest, NGINX, AWS, Docker

ML: Python, NumPy, pandas, sklearn, PyTorch, TensorFlow, Keras, Tensorflow serving, protobuf, Docker compose

Contact Us