Edinburgh Napier University

  AI hardware solutions are only useful if they’re compatible with all other layers of the technology stack, including the solutions and use cases in the services layer. Researchers in AI hardware can take two paths to achieve this goal. First, they could work with partners to develop AI hardware for adopting industry-specific use cases, such as oil and gas exploration, to create an end-to-end solution. Alternatively, it is to focus on developing AI hardware that enables broad, cross-industry solutions, as Nvidia does with GPUs.

However, researchers in AI hardware cannot rely on other hardware companies to build layers of the stack that will be compatible with their hardware. It is essential to create an ecosystem in which software developers/users can choose preferable hardware. In return, researchers ’ll have more influence over design choices in the industry. In this project, collaborating with Codeplay Ltd, I will develop an open software stack to draw software developers into the SYCL based ecosystem to reduce hardware complexity whenever possible. Since there are now more types of AI hardware than ever, including new accelerators, this project should offer simple interfaces and software-platform capabilities to allow software developers/users to take advantage of high performance hardware.