In this era of digital transformation, massive amounts of data are produced and collected. Business success depends on how well this data can be analyzed and the quality of insights produced. In the past, data scientists manually analyzed and drew inferences from relatively small amounts of data. As the amount of data produced expands exponentially, Artificial Intelligence (AI) and Machine Learning (ML) are key to enable data scientists to continually analyze the data and draw insights at never before seen levels. AI and ML are now used across a vast set of use cases, such as transaction processing, public safety and security, biomedical research, fraud detection, data indexing & search, supply chain optimization and many more.

Today, businesses have anywhere from a handful to thousands of ML models in production. To improve accuracy and insights, data scientists retrain their models on a weekly, daily and even hourly basis. With the growing number of use cases and data, the number of ML models generated is growing exponentially. This need to scale requires IT teams to invest in expensive AI/ML infrastructure, but at the same time it is a struggle to maximize utilization to justify the investment.

The Diamanti ML platform is the first bare metal enterprise platform with GPU support for running containerized AI/ML workloads under Kubernetes. The solution supports NVIDIA’s NVLink cross connect GPU card technology to maximize throughput of the GPU infrastructure. Atop the physical infrastructure, Kubeflow provides a machine learning framework for Kubernetes that deploys highly available Jupyter notebooks and ML pipelines.

                                                       Diamanti Machine Learning Platform

Diamanti Spektra can manage both GPU and non-GPU based resources in a single cluster or in a hybrid cloud setup. This provides the flexibility for data scientists to accelerate model training with GPUs and deploy CPU-optimized workloads to the same cluster. With Diamanti’s shared storage system, data is readily accessible for model training as well as to run analytics to draw insights. This avoids back and forth movement of massive datasets between the online processing system and off-the-cluster storage hence saving significant time for data scientists.

The Diamanti platform combines the power of Diamanti’s hardware offload, NVIDIA GPUs and lightning fast NVMe shared storage to provide a bare metal Kubernetes platform for data scientists to run ML workloads by maximizing resource utilization and return on investment.