Embedl is a product-oriented research spin-out with a strong focus on commercialising technologies and services to facilitate the uptake of AI and deep learning in strategic industry applications, e.g. autonomous driving, robotics and IoT. In VEDLIoT, Embedl continue on its mission to make deep learning efficient in embedded systems by developing new hardware-aware methods for deep learning compression. Moreover, Embedl develops novel methods for model-hardware co-design as well as automatic distribution and merging.
The methods for optimal distribution of a model will be exploited as an add-on to Embedl’s licence-based core offering, to enable up-selling, and pave the way for a modular pricing model, where new modules can be developed and charged separately from the core.
The model-hardware co-design methods in the project will be developed as proof-of-concept. Given promising results from this work, a collaboration consortium with potentially both project partners and external parties will be formed to define a new research project or commercial project.

