The Power Management and Multimarket (PMM) division at Infineon has a large sensor portfolio cherished by customers. Following a core strategy of Product to Systems (P2S), Infineon strongly believes the real value comes not only from delivering high-quality products, but by understanding better the customer needs and providing complete solutions. This approach significantly reduces the end customer's time to market and builds a better customer trust relationship, eventually leading to improved profit margins.
For developing smart solutions for sensors, Infineon leverages machine learning (specifically, deep learning) classification techniques in system designs. Example: Infineon's Alarm System. These machine learning classification algorithms are made available to customers as either part of the embedded solution, or offered as an online service.
For the latter, it is a huge challenge to set up an infrastructure in a scalable way that can run on-demand classification service requests coming from end customers. Luckily, MATLAB Production Server™ solves this problem by taking care of the deployment infrastructure while Infineon can focus on the algorithm development.
With MPS, an example web application using Machine Learning as a Service (MLaaS) deployed on Amazon Web Services (AWS) was successfully demonstrated. Apart from some networking issues due to the corporate firewall of the company, the deployment process was straightforward.
As a way forward, Infineon intends to use the demo setup for real applications; furthermore, they plan to investigate offering services like retraining the neural network on the cloud, targeting the embedded solutions.
