Client background & business context

Client: multinational corporation specializing in measuring instruments and metrological technology 

Software Product: Machine Learning-based desktop application capable to identify the defect detection on circuit boards as soon as they get off the production line. 


Great amount of manual work: until the application release, all defect detection had to be done manually, by an employee of the factory. This operation was prone to mistakes and demotivation due to the repetitive character of the tasks. 

Efficiency and proactivity concerns: circuit boards are usually small elements with details and defects spanning over a several pixels in an image. 

Non-stop production: 24/7 production requires human presence in the defect detection department. 

Our approach

We started the project with a 3-day Discovery Workshop, during which we mapped out the main features of the application, the general UI direction and way of working. Our chosen approach was Agile due to its benefits in dealing with changes and input throughout the course of the development. 

To make sure we are on the right track as soon as possible, we agreed on a Minimum Marketable Product version. We also established clear milestones, which established expectations and allowed us to deliver a first draft of the entire application flow. 

Delivery was done in 2-weeks increment, enabling both the team and the Business Product Manager to make the right decisions in terms of priority, dependencies and complex functionality research. Every increment had a Demo session planned, followed by an UAT period. 

The last month of the timeline was used for overall testing and quality assurance. We made sure that once the product was released on the market, it can operate in the established parameters. 


.NET Core 3  

.NET/ C#



Telerik controls


The end-product is a desktop application that allows users to prepare reference data sets of images, relevant for the circuit boards which need to be analyzed. The classification of proper vs defect images, followed by a marking of the defect areas, enables the Machine Learning algorithm to make a distinction and be able to identify by itself the defects, when applied on real-life production images. 

The software product allows our client to offer added value to its clients, by increasing production accuracy and efficiency. The application opens the door to new opportunities in measurements and quality assurance solutions, that make use of advanced technological approaches. 

Related case studies

Project Reporting Process Digitalisation

Cloud application to support the digitalisation and automa ...

Cloud-First Strategy: Microsoft Exchange to Office 365 Mig ...

On-premises Microsoft Exchange infrastructure migration to ...

API first integration for enhanced availability and reserv ...

Receiving notifications about any possible breaks in the ...


Mies-van-der-Rohe-Strasse 6, c/o Design Offices
80807 München, Germany
E-mail: hello (at)


Constanta 12, Platinia
400158 Cluj-Napoca, Romania
E-mail: hello (at)


Cetatii Square 1, Oradea Plaza
410520 Oradea, Romania
E-mail: hello (at)