HW/SW Co-design for Embedded AI

HW/SW Co-design for Embedded AI

KI

Context

AI plays a crucial role for all kinds of modern computing algorithm. Due to the compute intensive nature of the workload special accelerators are often used to enable a fast and energy-efficient execution. Similarly, the software stack plays a crucial role. It needs to strike a balance between ease of use and achievable performance.

Tasks

Your task is to further develop an AI accelerator. Base operations, such as simple layer computation, are working but more optimizations an new network types, such as transformers, are still needed to deliver sufficient performance. Therefore, your task will be to co-design new functionalities and a suitable software interface for them. Optionally, these can also be implemented into a compiler toolchain.

Requirements

  • Knowledge in Scala/Chisel in beneficial but not necessary
  • Knowledge in Python and C for software development
  • Interest in learning about AI in embedded domains