Machine learning infrastructure deployed at scale: understanding requirements, demand, impact and international best practice

Over the past two years, there has been a steep increase in the application of machine learning (ML) techniques across a wide range of domains. There has also been a significant increase in demand for hardware suited to deep learning, including GPUs and high-bandwidth file systems. This project investigated the soft and hard infrastructure required to underpin and support the increasing adoption of ML, at scale. Goals for the project were: 1. To form a clear understanding of the relationship between research requirements, computing capability, capacity, and research impact. 2. To understand individual researcher requirements and consolidate these across a large cohort of groups, so that we can make evidence-based recommendations on how to underpin research adopting ML in the most efficient and effective manner, at scale. 3. To understand international best practices, and how it should inform Australian investment