Case Study: Deakin University achieves 10x faster AI training with WEKA NeuralMesh

Deakin University achieves 10x faster AI training with WEKA

Deakin University's Applied Artificial Intelligence Initiative faced a significant challenge as its legacy storage infrastructure could not handle the I/O demands of modern AI research. This bottleneck, particularly for GPU-heavy workloads like computer vision and large language models, caused long delays in starting jobs and training models, slowing down critical research projects such as the development of AI for early cerebral palsy detection. The university partnered with vendor WEKA to overcome this barrier.

WEKA implemented its NeuralMesh high-performance storage solution, which consisted of 436TB of raw NVMe storage. This provided the exceptional I/O performance required, eliminating disk wait states and freeing researchers to focus on their science. The results were transformative: WEKA's solution delivered a 7 to 10x speed increase, reducing model training times from over an hour to just six minutes and accelerating the pace of breakthrough research.

Deakin University