Case Study | NovaLeah-AI-ML

Introduction

NovaLeah is a renowned cybersecurity firm specializing in the medical devices sector. Their mission is to provide robust cybersecurity solutions to ensure the safety and effectiveness of medical devices, protecting them from potential cyber threats. As part of their commitment to innovation and excellence, NovaLeah partnered with HST Solutions to develop an AI/ML-based software solution.

Project Overview and Challenge

The project was divided into two parts: MDS2 and MDS+. MDS2 aimed to assist medical device manufacturers in navigating the Manufacturer Disclosure Statement for Medical Device Security (MDS2) questionnaire, while MDS+ was developed to help medical device buyers create custom questionnaires for manufacturers.

The challenge lay in the complexity and technical nature of these questionnaires. They often contain hundreds of questions that require specific knowledge about the device’s safety and security features. Furthermore, the responses need to be consistent and accurate, as they are critical for ensuring the safety and security of the devices.

Manually answering these questionnaires can be time-consuming and prone to errors. Moreover, the responses can vary significantly depending on the person filling out the questionnaire, leading to inconsistencies.

Technology stack:

.Net Core | Azure Functions | Razor | Azure Data factory | Azure SQL | Azure containers | Miroservices

Services:

  • Product Strategy
  • Software Architecture, Analysis
    and Design
  • UI / UX Design
  • Born in the Cloud Software Stack
  • Full Life-Cycle Application
    Development
  • API Automation

AI/ML Solutions and Techniques

To address these challenges, HST Solutions developed an AI/ML-based solution that provides guidance and suggests appropriate responses to streamline and standardize the process of answering these questionnaires.

For MDS2, the software uses Fuzzy string matching, an AI technique that determines the similarity between different strings. This allows the software to match the device in question with similar devices whose questionnaires have already been answered. The software auto-answers these questions using historical data or provides recommendations based on responses from other manufacturers of similar devices..

For MDS+, the software uses a pre-trained semantic similarity model, distil-Roberta from Hugging Face Transformers, to perform semantic similarity searches. This allows buyers to search for questions similar to what they are looking for, enhancing the customization of their questionnaires.

Technical Implementation

The software was deployed using Azure Functions, a serverless computing service that allows for automatic scaling and efficient resource management. The MDS2 and MDS+ products involved a total of nine micro-services, each designed to perform specific tasks such as searching for devices or questions, building and reviewing questionnaires, and suggesting or upserting answers.

The backend was built using a No-SQL database server to handle the unstructured data common in NLP tasks. This database was designed with specific keys to allow unique identification of contents, ensuring efficient data retrieval.

To handle the ‘cold start’ problem common in serverless architectures, different session-state handling tools were used to prevent delays in the product.

The AI/ML models were optimized to work within the constraints of the serverless environment, ensuring they were as compact as possible without compromising their performance. The models were also designed to be stateless to fit with the serverless architecture, with any necessary state information stored in the database.

Overall, the technical implementation of this project demonstrates a careful balance of advanced AI/ML techniques, efficient software architecture, and innovative problem-solving.

Conclusion

The partnership between NovaLeah and HST Solutions resulted in a cutting-edge AI/ML software that significantly enhances the process of creating and answering questionnaires for medical device manufacturers and buyers. Despite the technical challenges, the project was successful in leveraging advanced AI/ML techniques and serverless architecture to deliver an effective and efficient solution. This project further solidifies NovaLeah’s commitment to leveraging advanced technologies to improve cybersecurity in the medical devices sector.

Technology stack:

.Net Core | Azure Functions | Razor | Azure Data factory | Azure SQL | Azure containers | Miroservices

Services:

  • Product Strategy
  • Software Architecture, Analysis
    and Design
  • UI / UX Design
  • Born in the Cloud Software Stack
  • Full Life-Cycle Application
    Development
  • API Automation

Contact Us

Tell us about your custom software project.

Let our team

Be your team

Unsure of how to get started? Talk to us. No matter where you are in the process, we can help you. Whether you need us to design and build a prototype, take your project from concept to maturity, or build off the work of another team, we can make your idea a reality. We also step in when it makes more sense to lean on our expertise than to have an in-house team get spread too thin.

Generally, we are able to respond to inquiries within 8 business hours.

Please fill in the form below and we will be in touch.








    What is 7 + 5 ?