We use cookies to ensure you have the best possible experience. If you click accept, you agree to this use. For more information, please see our privacy policy.
/

KonnectAi

Scaling AI for smarter manufacturing inspections
Close up of hand using the Konnect Ai app to test an electrical wire.

The context

Our client, Promark Electronics, a Division of Electrical Components International (ECI), is one of the world’s leading suppliers of wire harnesses, electrical distribution systems, and other critical engineered components for industries such as aerospace, medical, defense, and transportation. With 25,000 employees and 40 global manufacturing locations, ECI has become the trusted partner to market leaders with over 500 customers around the world.

Since their inception, Promark’s visual inspection has been done manually, with some inspections taking up to several minutes per part. On a mission to drastically reduce this time, incorporate advanced technology, and ensure component traceability, they came to us seeking development of an innovative system called KonnectAi . Employing state-of-the-art AI algorithms and computer vision technology, this solution would facilitate in- process and final visual quality inspection. Here’s how we did it.

Project details

Industry
Manufacturing
Technologies
Google Cloud Platform
Computer Vision
Go
Python
ML Ops
Serverless Architecture
CI/CD
Services
AI
Design
Development

The challenge

Promark conceived the idea of visually inspecting parts using a camera encased in a 3D printed station. An operator on the production floor would position a component to be inspected, such as a cable, under the camera, which would stream video to a connected device. A sophisticated computer vision model would then analyze the component and determine its acceptability based on various criteria. The AI model would scrutinize key aspects of the part (e.g. whether the screws are properly in place or whether the end connections are correctly executed) to ensure each part meets the manufacturer's strict quality standards.

If the component is deemed “acceptable” it would be recorded in the system as such and delivered to the customer. If the component is deemed “unacceptable,” the operator would then manually inspect the item, either confirming the defect or flagging a false positive in the system. Timestamped photos of each part would be added to the database for complete traceability, should that component ever fail down the line.

Promark wanted KonnectAi to have five key sections. The Wizard acts as the interface for Promark production employees to create new inspection points by submitting pictures of parts for AI analysis. This section would also train the AI model to detect issues accurately. Vision utilizes the camera to identify defects in parts, employing AI for precise defect analysis. The Inspector serves as the validation interface for final quality inspectors, enabling them to verify part quality with data integrated from Vision. This ensures rigorous quality control, allowing users to manually review inventory and inspect for false positives among parts. The Client Admin Panel is designed for individual clients to manage inspection points, requests, users, and roles, equipped with a dashboard that offers strategic insights. Lastly, the Super Admin Panel provides overarching system control, model management, and the ability to review feedback and communicate directly, facilitating operations across all clients.

Beyond developing KonnetAi's models, a central challenge involves their maintenance and scaling once they're in production. Whereas the realm of machine learning (ML) focuses on creating models that can learn autonomously, Machine Learning Operations (ML Ops) emphasizes the automation and routine training of models to prevent them from becoming less effective as production variables change over time. Our task was to develop a comprehensive pipeline encompassing data collection, managed by the Wizard, through to automatic ML training facilitated by Google Cloud infrastructure, and culminating in deployment via continuous integration and continuous deployment (CI/CD) processes.

Close up of electrical wires.

The task at hand

During the initial phase of our project, we embarked on a discovery process, addressing the UX and UI aspects to ensure an optimal user experience. We also provided the necessary technological support and training, aiming for Promark to eventually manage the software independently.

Delivering high-performance AI solutions without substantial cost is always a challenge. The allure of AI technology is undeniable, yet the associated costs can be somewhat elusive. And so, we worked to develop a solution that not only costs a fraction of what a company would incur with Google Cloud's plug-and-play services, but also offers greater control over data management and system customization. While using out-of-the-box cloud solutions offer quick setup, our approach required more time to refine, granting us greater flexibility in data handling and system adjustment. To manage data storage more efficiently, taking into account the potential volume of video streams, our system prevents the storage of duplicate videos, maintaining an organized and cost-effective cloud environment.

On the technical front, scaling from a single AI model to dozens without escalating costs or complicating infrastructure required creative solutions. Our discussions with Google led us to adopt a serverless architecture, which was both cost-effective and efficient. By transitioning our code from Python to Go, we achieved a significant boost in processing speed, enabling smoother cloud functionality and the ability to manage an increased number of AI models effectively.

Data protection was another critical area we addressed. We opted for a Software as a Service (SaaS) model, where each client operates within their own Google Cloud project adhering to strict Identity and Access Management (IAM) policies, ensuring data privacy and security through strict identity and access management policies.

Close up of hand using the Konnect Ai app to test an electrical wire.
Mockups of Konnect Ai application screens

If the component is deemed “acceptable” it would be recorded in the system as such and delivered to the customer. If the component is deemed “unacceptable,” the operator would then manually inspect the item, either confirming the defect or flagging a false positive in the system. Timestamped photos of each part would be added to the database for complete traceability, should that component ever fail down the line.

Multiple Konnect Ai desktop screens.

The end result

We started working with Promark in the Summer of 2023 and are pleased to have delivered V1 of the product in December 2023. This advanced solution now equips Promark to conduct thorough inspections, leveraging computer vision to serve as an extra pair of eyes to never miss a detail, track and record everything to ensure top-notch quality, especially for must-never-fail (MNF) products. KonnectAi is a significant breakthrough in the electrical distribution systems sector.

Each AI model is uniquely designed for specific component analysis, with the potential for scalability to hundreds of models. This scalability is supported by KonnectAi's continuous learning and dynamic image analysis capabilities.

What’s next

Up next, we’re developing the client admin panel and the super admin panel for Promark. We’re implementing tracking analytics and adding many new mechanisms to make sure the solution is scalable for different clients. And new clients can get up and running with their first model in a single day! We’re on track to wrap up the project in June 2024.

This project was a great experience in scaling and building cost-effective AI models. We look forward to working on more AI-powered projects like this one, that deliver sustainable, powerful results over the long-term. If you’re a company that needs help with AI, or if you already have an AI expert on your team and have already built a model, but need help figuring out what comes next, get in touch.

Konnect Ai screen presented on a tablet.

Insights

It starts with an idea
Culture

Osedea portraits: AI, adventure and culture with Isabelle Bouchard

Marie-Pier Houle
Marie-Pier Houle
read time
Development

A minimalist approach to DataOps and MLOps with DVC and CML

Nicholas Nadeau
Nicholas Nadeau
read time
Culture

How we run the show at OSEDEA (without any bosses)

Ivana Markovic
Ivana Markovic
read time
Development

Capitalizing on your organization’s data with vector databases

Carl Lapierre
Carl Lapierre
read time
Culture

Elevating engagement in hybrid workspaces

Ivana Markovic
Ivana Markovic
read time
News/Press

Introducing Osedea's equitable and inclusive parental program

Ivana Markovic
Ivana Markovic
read time

We’re on a constant quest to broaden our horizons and spread wisdom. It’s all about pushing boundaries and elevating our game.

Read more insights
Button Arrow