Many organizations rely on pretrained models from ISVs to generate operational insights and real-time alerts from computer vision and AI software. However, other applications of machine learning that are specific to unique use cases are often needed. Meeting these needs requires a combination of an easy-to-manage, GPU-enabled hardware platform coupled with a full-feature MLOps platform. The Dell Validated Design for Computer Vision platform leverages a VxRail HCI cluster with VMware virtualization for both VMs with VMware vSphere and containers with TKG. The VMware platform and tools are familiar to many IT professionals worldwide and provide a cost-effective common platform for adding containerized applications into new and existing environments. The BOSS AI data science framework provides a full set of data science product development and management features that will increase the productivity of any machine learning team efforts to develop new ML/AI applications.
The solution provides two important classes of benefits:
- Infrastructure simplification capabilities
- Data science toolset hosting capabilities
Infrastructures
The benefits of the platform are:
- Many CV vendors are adding container-based application offerings that benefit from a standard -based environment for hosting.
- The ability to perform any container-based workloads alongside CV workloads hosted with VMware virtual machines simplifies the management of complex systems.
- Hosting containerized CV workloads that require GPUs using a previously validated TKG solution on the VxRail platform if supported by the developer
- Supporting multiple isolated clusters with TKG to allow running any container-based workload alongside a suite of CV applications on a single platform.
Data science capabilities
The BOSS AI ecosystem provides the following benefits:
- Running the BOSS AI application alongside the VMS and CV systems allows access to data without leaving the cluster. This ensures data is not transferred to a remote location to perform analysis and avoids compliance issues due to conflicts with the many types of sensitive data requirements and certifications, including:
- HIPAA
- PCI-DSS/PHI/ PII
- SOX
- HITRUST HIGH
- ICD 503 (classified data)
- GDPR[MC1] /CCPA/CCA/LGPD/ POPI
- SOC 2, ISO27001
- An AI model can be trained with a "No code" interface that lowers the barrier to entry for a team. Full customization is also supported for advanced users.