Futures: Custom AI Hardware Platforms

“Custom AI Hardware Platforms” provides a detailed exploration of custom Artificial Intelligence (AI) and Machine Learning (ML) hardware implementations, starting with an introduction to Field Programmable Gate Arrays (FPGAs) and the synthesis of digital design implementations to these reconfigurable devices. Then, implementations of ML algorithms, such as deep neural networks, on FPGAs will be presented and analyzed for their efficient and accelerated processing power-saving capabilities. The course then introduces the Application Specific Integrated Circuit (ASIC) and explores the design pipeline from prototyping hardware-based ML algorithms with FPGAs to production in ASICs. Then, neuromorphic computing, the technology inspired by the human brain, will be introduced to examine a system view of the combination of hardware and software components in the implementation of AI tasks and ML algorithms. The course, and series, concludes with a Capstone project, where students will apply their knowledge by implementing an AI/ML algorithm on custom hardware, gaining hands-on experience in designing, optimizing, and deploying system-level (hardware and software) solutions.