Enhancing IoT Device Management Through AI-Driven Solution

The Customer

Our client, an IoT platform hosting 100k of digital twins across more than 200 types, faced a significant challenge with their web interface. Users struggled with a difficult process to find and configure settings for each device type. The lack of usability led to operational inefficiencies and increased costs, as users spent excessive time navigating the complex interface.

The Challenges

• Online system with thousands of active users;

• Wide majority of types of devices;

• Requirements to make type detection almost immediate and the most accurate.

The Solution

To address this, we assembled a specialized AI team consisting of a business analyst, a data scientist, a machine learning engineer, front-end and back-end engineers. Our approach was methodical and client-centric:

​

1. In-Depth Analysis: Our business analyst first delved deep into understanding the problem. Through thorough analysis, the complexity of the user interface and corresponding operational challenges were better described and documented.

2. Project Planning: Collaborating closely with the client, our team proposed a typification algorithm, aimed at automatic detection of device's types and then simplifying the interface based on the algorithm's work results. This solution was designed to make the configuration process more intuitive and efficient by adopting a user interface as fast as possible right after adding a new device to the system.

3. Data Validation and Preparation: We proactively initiated a data validation process. This critical step was taken to ensure the feasibility of solving the problem at hand with the available data resources. Our commitment to data validation underscored our dedication to making informed, data-driven decisions throughout the project. After confirming that the current data is suitable for the solution, the data scientist and machine learning engineer worked on collecting and processing vast amounts of data to prepare it for the machine learning solution.

4. Machine Learning Implementation: We developed and trained a machine learning model to recognize and categorize different device types accurately. Achieving an F1 score of 88% indicated high precision and reliability of the model. Fast speed was implemented with help of pre-configuration functionality.

5. User Interface Update: Concurrently, our front-end engineers redesigned the user interface. The new design showcased configurations relevant to the detected device type, enhancing usability while retaining the option to manually change device types if needed.

6. Seamless Integration: The updated system was integrated into the production environment in close cooperation with the client's engineering team under support of our back-end engineer. Canary Deployment scheme was used to ensure a smooth transition without any downtime and ensuring that there will be no surprises or any other unpredicted behaviour.

The Outcome

Our solution revolutionized the client's IoT platform in several ways:

• Reduced Configuration Time: The time spent on digital twin configurations was cut by 48%, significantly reducing operational burdens.

• Increased Device Effectiveness: The effectiveness of device usage saw a 36% increase, thanks to the more intuitive configuration process which helped to make devices bring value significantly faster.

• Enhanced Client Loyalty: Perhaps most impressively, there was a 65% increase in client loyalty, underscoring the substantial improvement in user experience and satisfaction.

Conclusion

This case study demonstrates the power of a well-thought-out solution combining machine learning and user interface design to address a specific operational challenge. Our team's strategy and AI with IoT technical expertise not only streamlined the configuration process but also led to significant improvements in operational efficiency and customer satisfaction for our client.

Technology Stack

Front-End Development: React.js and JavaScript

​

Back-End Development: Python

​

AI and Machine Learning: TensorFlow and Pandas and NumPy
 

Database: PostgreSQL 
 

Cloud Platform: AWS
 

DevOps and Deployment: Docker and Kubernetes