Black Pepper came in to our business and took the time to understand our data and business domain first, in order to provide useful Machine Learning software.
Jaguar Land Rover (JLR) is a leading global automotive manufacturer that places quality at the heart of its philosophy. In order to manage its quality process, JLR has made significant investment in the development of its product monitoring systems, which provide valuable feedback to the business to ensure quality remains at the heart of its ongoing development.
To streamline and optimise the current processes, JLR partnered with us to conduct a discovery experiment around data that they had collected over many years to assist in a long term vision of intelligent quality assurance.
We utilised our artificial intelligence knowledge and machine learning expertise to investigate whether the data provided was amenable to machine learning. The detailed discovery period delivered findings iteratively and in a timely fashion. Our previous engagement with JLR allowed the development team to understand the business domain quickly and comprehensively to ensure the end result was high quality and focused on the client needs.
The data discovery allowed us to fully understand the business before using resource on development.
Based on the specificity of the sector, and JLR's area of expertise, a set of technologies were chosen to make the solution as specific to the client as possible. This was especially important to the customer as previous attempts to optimise their data pipelines have proved to be a challenge.
To ensure our solution was entirely suited to the business case, the use of a domain-specific dictionary, custom architecture was employed, alongside a deep understanding of JLR's current internal processes. Where possible, the development cycle involved automated optimisation of the machine learning system in order to focus resource on alternate approaches that may also provide value.
Throughout the discovery period, it was noted that while JLR's data was recorded extremely well, it was not entirely uniform and presented a handful of difficulties. We understood that this is expected in a real-world business problem, and used several methods to counteract 'natural' data spread issues.
JLR is now in a position to move towards a machine learning pipeline to optimise the flow of user feedback to the correct respective departments. The proof of concept system we developed demonstrated that the business could benefit from adopting artificial intelligence mechanisms, allowing their employees to work in conjunction with any new systems to provide business value in the most difficult areas.
By working closely with the customer and working on the good relationship previously built between the companies, we were able to understand the business at an employee level and improve the first iteration's machine learning system accuracy from 45.8% to 84%. Our proof of concept has evidenced that Machine Learning could play a key role in JLR’s quality control department, not only improving customer satisfaction and increasing efficiencies but also supporting their innovative aspirations.
The data discovery output provides a platform for JLR to build upon, and offers the flexibility to support a multitude of other data feeds that JLR have collected in parallel.