Video

Make AI Accessible for Everyone

Effortless, scalable AI deployment with Chariot and Striveworks partners—automating MLOps, reducing costs, and enabling innovation across any infrastructure.

Google Cloud C3D VMs Powered by AMD Make AI Accessible for Everyone (Featuring Striveworks)

 

Transcript
(Automated transcription)
 
Jay Marshall: Lots of people talk about democratizing AI, but you can't democratize AI if you stop at the hardware. And lots of folks are starting to realize that deploying AI in production, at scale, can be challenging and very expensive if they wait until the end to consider their hardware options.
So things such as provisioning: compute, setting up GPU device drivers, managing model checkpoints and artifacts, managing data sets. And it really just should be taken care of for them and automated.
And that's what our platform, Chariot, aims to do. So really just allowing data scientists to focus on data science.
Now with this stack, in Chariot, our customers can train, deploy, and monitor models on a very low-code, no-code way. And that deployment piece is taken care of by NeuralMagic and C3D.
So models are running on CPU, especially with the C3D instances compared to the earlier generations. You're not sacrificing any speed by moving those models off of GPU.
 
And so that really just saves our customers money, allows them to train more models, it's
Jay Marshall: At its core, NeuralMagic is an ML optimization company. We help customers optimize their machine learning models and then run those models as performantly as possible on the underlying hardware.
 
Now the latest Google Cloud C3D instances, powered by AMD EPYC™ processors, bring even more performance than the prior generation N2D instances at, ultimately, a lower cost.
And Google Cloud's own internal testing found 2x performance gains on NLP models like BERT and up to 3x performance gains on computer vision models like YOLO and ResNet in gen-over-gen performance comparisons.
on Google Kubernetes Engine, or even on physical hardware, like Google Distributed Cloud.
But by bringing this kind of performance to CPU-based architectures like the C3D instances on Google Cloud, NeuralMagic is giving customers choice by deploying on the ubiquitous x86 processors that they already know how to manage.
This choice, and the operational efficiencies that come with it, move us one step closer to the true democratization of AI.

Video Summary

The Striveworks MLOps platform, Chariot, enables seamless model deployment across any infrastructure. By connecting Neural Magic DeepSparse, Google Cloud C3D instances, and AMD EPYCTM processors, data science teams can run AI models in production on CPUs, reducing cost without sacrificing performance.

This is another way Striveworks lets you bring the platform to your data, whether you operate in cloud, multi-cloud, hybrid, or on-prem.

The Striveworks MLOps Ecosystem

Striveworks has relationships and deep integrations with many of the best organizations in the technology landscape. See our MLOps ecosystem. 

Related Resources

Striveworks partners with Neural Magic

Striveworks Partners with Neural Magic for GPU-less Model Deployment

Integration of Neural Magic's SparseML and DeepSparse into Striveworks' Chariot allows customers to deploy models using only CPUs while getting GPU-like speeds.

Read Press Release
Striveworks' Eric Korman and Neural Magic's Jay Marshall appear on Neural Bits episode on Disappearing MLOps

Is MLOps Disappearing?

Striveworks CSO, Eric Korman, explains "disappearing MLOps" and how, together, Striveworks and Neural Magic can help customers accelerate their ML lifecycle for faster business impact.

Watch Video
AMDs Advancing AI 2024

Striveworks and Neural Magic Mentioned in AMD's Advancing AI 2024

Striveworks and Neural Magic are running inference workloads on CPUs, saving money without sacrificing speed.

Watch on Youtube