How we streamlined a manufacturing inspection process using an Amazon Rekognition solution

Overcome the high cost of scaling PCB inspection automation

Our client developed a highly effective electronics manufacturing software platform. So effective, it demanded substantial growth. However, the cost of automating PCB inspections on their existing Google Cloud Platform solution was prohibitively high for the scale they desired. To find a more cost-effective and robust solution, they turned to zeb and our deep expertise in Amazon Web Services (AWS).

Leverage advanced AI and image analysis for efficient inspection

As an Advanced Tier AWS Consulting Partner, zeb put our experience to work, rapidly developing and deploying a custom AI model aligned with our client's business objectives. Our team leveraged the power of Amazon Rekognition, an advanced vision platform that harnesses the deep learning capabilities of generative AI. This approach provided an efficient and effective pathway, allowing our client to scale without technology as a hinderance.

Efficiency, accuracy, and adaptability

Outstanding Cost Efficiency: Our solution reduced our client's previous cost of inspecting a single PCB component using their Google Cloud solution by an incredible 99%. These savings will continue to grow with higher volumes, thanks to Rekognition's low hourly inference rate of just $4. Training and scaling the model are equally affordable, at only $1 per hour.

Exceptional Accuracy: Our solution not only slashed costs but also delivered outstanding performance, achieving an impressive 99.66% accuracy in detecting components on PCB boards. This high level of precision ensures reliable quality control.

Rapid Development and Implementation: zeb designed and developed this innovative solution in just one week. While not every solution can come this swiftly, our team's ability to quickly understand a client's needs and deliver a high-quality, effective solution is one of our greatest strengths.

Efficient Model Management: The solution we built is highly adaptable, featuring a robust system for model management that allows for easy updates. With a new dataset uploaded to Amazon S3 and a simple API trigger, the model can automatically update itself with the latest data. Additionally, Amazon Rekognition can be started or stopped as needed, giving our client complete control.

Organized Data Management: A versioning pipeline preserves previous datasets whenever new ones are added to update the model, ensuring data is never lost. This organized approach to data management makes it easier to track changes and maintain high standards of quality.

Transforming your manufacturing operations with technology is easier than you think. As your strategic partner, our team will assess your specific needs and objectives and tailor the right solution using our expertise and proprietary methodology. Let's have a conversation – contact us below.