Embedded KI: Scalable, integrable, maintainable

Hardware design as a basis

Contrary to the classic approaches to model development, we recommend that our customers pre-select the embedded AI hardware in good time. The aim is to project the footprint of the embedded AI hardware onto the business case (cost impact). Nevertheless, our approach retains the flexibility for various AI extensions.

Of course, model development remains the central element at the beginning. Whether with your own model developer or with an expert from our environment - we start with the conception in order to select the software environment as well as the possible target hardware.

Different software toolchains: Integration into the target hardware

Depending on the existing development environment for your model, the transfer to the target hardware must be secured. This may require the integration of runtime environments, or an open data exchange format such as ONNX is used to make the model executable on the target hardware.

Project management «AI conformal»

The mixture of agile model and software development and classic hardware development takes place in a well thought-out project planning approach. AI lifecycle management is integrated into project management as a new ‘discipline’. Because your AI model is ‘alive’: to ensure good AI results, the algorithms of your model are ideally monitored, trained and optimised on an ongoing basis.

Our deployment tool GUT! (Grossenbacher Update Tool) is the basis for your diagnostic and update requirements and can be customised to your specific needs.

Holistic approach for your project success

Our approach to embedded AI focusses on the scalability of the hardware. This is to achieve the ideal mix in terms of costs and performance. But also to make an existing basic product, which is to be expanded with embedded AI, manageable in a first step and thus be able to manage the lifecycle of the standard product separately.