Increasing Packaging Efficiency with AWS SageMaker for a Pharmaceutical Company

Customer Profile
North America
3 Months
Technologies Used
AWS SageMaker
AWS Rekognition
Amazon CloudWatch
AWS Lambda
Key Outcomes
Accuracy in detecting
defective packaging
Decrease in drug rejection
Overhead costs reduction

Problem Statement

Pharmaceutical manufacturers produce a variety of drugs every day. The industry has strict guidelines for packaging and quality assurance. Once the tablets or capsules are manufactured, they are packed in blister packs using blister packaging machines.
As with any production line, pharmaceutical manufacturers face a high drug rejection rate due to problems related to damaged containers, missing tablets in blister packs, and packaging of broken tablets.
The rejected drugs increase the overhead costs as they are checked during shipping. To tackle these problems, assessing the packages while still in the production cycle is crucial.
The existing process of identifying defected containers relied on manual checks at various stages of the production cycle, which required dedicated personnel and prone to errors. Some of the prevalent issues included:
  • Damaged / Crushed blister packs
  • Missing labels
  • Shape and size alterations
The client was using high-definition cameras and capturing the visual information but was having difficulty in extracting valuable insights.
Collectively, these were the main concerns for our client. The domino effect caused by faulty blister packs was affecting their profits, brand reputation and customer satisfaction.


Automated Quality Assurance Using Computer Vision
With new and innovative technologies, quality checks can not only be automated but also result in higher accuracy.
Visual analysis using artificial intelligence combines learning and experience of how humans learn from visual inspection. The most transformational factor is the speed at which computer systems can process the data.
Upon analyzing the customers’ requirements, our experts identified the core areas where automation and machine learning could help.
Due to the clients’ existing infrastructure, we went with AWS SageMaker to process the camera feed of the packaging process and then develop a machine learning model from the visual data that detects damaged blisters with higher accuracy.
With clearly defined objectives, our team set out a strategy and defined periodic KPIs to ensure the project goes as planned.
Our AWS experts:
  • Designed a custom machine learning model based on Computer Vision
  • Trained the model with various images of the drugs, packages, and containers
The model analyzes every package & assigns a label based on:
  • Quality of the packaging
  • Tablet specifications – size, form, structure
  • Number of tablets in every package
This analysis is further sent to the sorting machine where it rejects the package from the production line eliminating the possibility of defective drugs making it to shipping.
The collected data, over a period of time, will enable the client to create machine learning models which could explain the reasons behind defective packaging.
Along with the project delivery, we also ensured that the in-house team gets all the knowledge to manage, monitor, and govern the model. Our experts conducted a tailor-made UX workshop to ensure that the end-users get the best experience with the solution.
  • Increased defect detection rate
  • Identifying packaging issues before shipping
  • Lowered drug rejection rate boosts sales
  • Optimized labor costs
A pharmaceutical manufacturer transformed their quality assurance testing with Attri’s expertise. As an AWS Sagemaker Consultant, we implemented a visual analytics solution based on machine learning that detects damaged blister packages before they are sent for shipping and distribution. The solution greatly increases detection accuracy and reduces drug rejection costs.

Want to build a custom machine learning solution for your business?

Talk to our Sagemaker Consultant
arrow icon