The Canonical Stack for Machine Learning will be fueled by easy access to efficient GPU capacity

The evolution of the Canonical Stack for Machine Learning will be fueled by easy access to efficient GPU capacity.

The AI Infrastructure Alliance is creating a framework in which the "Cambrian explosion" of Machine Learning Operations ("MLOps") technologies can evolve toward a "Canonical Stack" for ML — where the essential building blocks to drive AI applications are widely understood and accessible.

Founder Daniel Jeffries does a much more interesting job of explaining what the AIIA is, why it was founded, and why it's important.

Why does Juice belong in AIIA?

MLOps is a busy, dynamic space. Tools, technologies, and platforms are evolving and expanding rapidly as startups build, partner, integrate, and generally position themselves in this emerging landscape.

Most of the partners in AIIA cover parts of this landscape, as they build out their technologies to specialize in areas of coverage across Notebooks, Dashboards, Code Repos, Data Engineering Orchestration, Data Ingestion and Transformation, Data Synthesis, Experimentation, Training, Model Validation, Deployment, Production Inference Services, Monitoring, etc. — the emerging Canonical Stack of ML.

At first glance, it might seem that Juice doesn’t quite belong. Our technology operates on a fundamentally lower level than these ML applications, services, and platforms.

However, the very fact that Juice operates at a low level — making access to GPU easier and far more efficient — makes Juice a key enabling technology across many regions of the Canonical Stack, which will make it possible for other AIIA partners to build their platforms, and operate them for their customers, with greater ease and at substantially lower cost.

Where we fit in (and at the bottom edge of) the emerging Canonical Stack

Juice transforms how GPU Infrastructure supplies acceleration capacity to ML Experimentation, Training, and Serving Engines — like this:

‍

‍

As we rapidly add breadth to our core Remote GPU (rGPU) tech, supporting operating environments and APIs that are in widespread use within MLOps, we’re excited to be working with AIIA and its other great partners to contribute to the amazing vision of the Canonical Stack.

Onward and upward!

‍