CONFIDENTIAL designator

NVIDIA & Red Hat for AI

based FSI services

March 2023

1
 Registration link

March 20-23 | www.nvidia.com/gtc

 NVIDIA Leads in AI
Providing the Frameworks and Platforms for Development and Deployment

Transforming industries

1B 250 3,000 35K COMPUTER

VISION
CONVERSATIONAL
AI
RECOMMENDER
CUDA GPUs ExaFLOPS GPU Apps Companies Run S
in the Cloud NVIDIA AI

HEALTHCARE SCIENTIFIC AI ROBOTICS

450+ 12K 3.5M

SDKs +
AI Startups Developers
AI Models

AUTONOMOUS
GAMING VEHICLES AR / VR

Breakthroughs in deep learning around 2012 brought AI into focus, but only NVIDIA had the strategy, vision,
and roadmap to invest in supporting these now mainstream AI workloads,”
Forrester Wave, AI Infrastructure, Q4 2021
Large Language Models
Go Way Beyond Words
Large language models are the most
important AI models today. The engines
behind generative AI and the burgeoning
text-to-image medium (example pictured),
LLMs are also driving recommender systems,
conversational AI, drug discovery, and
possibilities not yet imagined.

NVIDIA Hopper, with its Transformer Engine,

will make large language models accessible to
everyone.

Image created by Midjourney based solely on the

text input; "Artificial intelligence, Leonardo da
Vinci drawing style"
 Accelerating digital transformation in Finance
AI/ML Optimizes Performance and Outcomes

BANKING CAPITAL MARKETS INSURANCE PAYMENTS FINTECH

Underwriting AML/KYC, Fraud HPC Pricing & Risk Algorithmic Trading

AI Intelligent Automation Customer Service Personalization Virtualization

5
Note: For more information, Refer to FSI Use cases in the Appendix leveraging GPU
NVIDIA AI-Ready Platform for
Red Hat
 NVIDIA AI Enterprise with Red Hat OpenShift
End-to-end open platform for production AI

Text Process Conversational Image AI/ML Existing

Recognition Automation AI Analytics Applications

NVIDIA AI Enterprise
AI and Data Science Cloud-Native Infrastructure
Tools and Frameworks Deployment Optimization

Data Scientist/
Developer/
AI Researcher

Self-Service Access to Securely Automate Kubernetes-Powered

Infrastructure Resources MLOps Pipelines Application Development
IT Administrator
MLOps
Mainstream Servers Public Cloud

NVIDIA GPU NVIDIA DPU CPU-only

 NVIDIA End-to-End AI software suite
Deep Learning Streamlined From Conception to Production at Scale

DATA TRAIN AT OPTIMIZED FOR DEPLOY AT

PREP SCALE INFERENCE SCALE

TAO
TOOLKIT

• Reduces data science processes • Train, Adapt, Optimize Models in • Maximize throughput for latency- • Fast & scalable AI to applications
from hours to seconds hours vs. months critical apps w/ compiler &
• Diverse query types – real time,
runtime
• 70x faster performance than • Open-source ML frameworks offline batch, ensembles
similar CPU configuration optimized for GPU • Optimize every network (CNNs,
• Up to 266x performance increase
RNNs, & Transformers)
• 20x more cost-effective than • Integrated w/ NVIDIA RAPIDS to over CPU-only
similar CPU configuration simplify development • Optimizes use of GPU memory
• Triton with FIL backend delivers
bandwidth
best inference performance for
tree-based models on GPUs
AI Adopters Are Keen on Using Containers To Improve AI Workloads
Benefits of Containers for AI adopters

94%
of AI adopters are
using or plan to
use
containers within
one year 1

Q: Are you using a containerized environment to deploy machine learning models to production? (n=493)
Q: What benefits [do or will] containerized environments bring to your AI initiatives? Please select up to 3. (n=638)
Base: ML is in production or proof of concept, or there are plans to use ML in next 12 months.1
Source: 451 Research, part of S&P Global Market Intelligence – Voice of the Enterprise: AI & Machine Learning, Infrastructure 2020
 NVIDIA AI Enterprise Certified on Red Hat OpenShift
Unleashing AI on Every Enterprise

• Streamline deployment and management of

AI, ML and data analytics workloads
consistently across hybrid cloud

• Securely deploy an end-to-end AI solution

with complete confidence that can also be
folded into an existing enterprise
infrastructure

• Facilitates agility, innovation, and rapid

delivery with self-service data science and
integrated DevOps/MLOps
10
Red Hat and NVIDIA collaboration
NVIDIA GPU Operator
 From Datacenter to edge
Typical Enterprise Workflow

Training Data

OPENSHIFT GPU OPERATOR NGC REGISTRY

Consistent deployment across platforms Automate GPU resources in Kubernetes NVIDIA AI

Simplify multi-container deployments Installs drivers and monitoring tools Industry application frameworks
Auto scaling and self healing Single golden image for IT to manage Partner applications
 Red Hat and NVIDIA enable Enterprise-Ready ML architecture

Set Gather and Develop ML model Deploy ML models Implement ML models

goals prepare data in app dev process apps & inference monitoring &
management

ML/DL and DevOps tools

ML data sources: databases (e.g., SQL, NoSQL), data lake, etc.

Hybrid and multi-cloud platform with self-service capabilities

Compute acceleration (GPU, DPU)

Infrastructure
 AI Platform for AI Driven Finance

AI
BUSINESS INFRASTRUCTURE AI PLATFORM FOR FINANCE OUTCOMES
DRIVERS CHALLENGES
Fraud AI Financial
AI Adoption path:
Detection • Build your own
Applications • Strategic Partners ALL END
Virtual USERS
Assistants AI Infrastructure Time to Insight
Complexity Data Science/
Data
Scientist /
Cross-Sell MLOps Researcher

System
Next Best Admin/Dev
Action Platform Ops
Scalability & Productivity at Scale
Productivity Core
Credit Default Lines of
Business
Prediction
Marketing,
Compliance Compute, Networking & Storage Finance,
etc.
Automation Total Cost of Hardware accelerators Return on
Ownership Investment

cited their struggle with how to

84% fear missing their growth objectives
if they don’t scale AI1 76% scale AI across their business1

14
1 Accenture: "AI: Built to Scale, From Experimental to Exponential." 2019.
 Getting Started With NVIDIA AI and Red Hat OpenShift
NVIDIA AI Enterprise Trial Programs

APPLY NOW FOR FREE NVIDIA LaunchPad Evaluation Software

TO GET STARTED • AI development and deployment trial • Requirements: NVIDIA-Certified
program System
• Deep dive, hands-on labs for AI • Free evaluation licenses for on premises
practitioners and IT staff POC
• Requires ~8 hrs / Access for 2 wks • 90 days to test and experience
 NVIDIA BlueField Data Processing Unit
TRADITIONAL SERVER DPU ACCELERATED SERVER

CPU CPU

VMs Containers

Infrastructure Management Software-defined Security

Software-defined Storage Software-defined Networking VMs Containers

NVIDIA ConnectX SmartNIC NVIDIA BlueField DPU with Arm Cores & Accelerators

Acceleration Engines Infrastructure Management Software-defined Security

Software-defined Storage Software-defined Networking

Manual Infrastructure Management | Security Appliances |
Storage Systems | Static Networks | Microservices |
East-West Traffic | Storage Access | Zero Trust Security Acceleration Engines

Offload | Accelerate | Isolate

16
 Disaggregated and Composable, Cloud-Native Compute Platform
Red Hat OpenShift on NVIDIA BlueField DPUs roadmap, isolated Data Center Control Plane

Tenant Cluster

x86 x86 x86

Red Hat PCIe PCIe PCIe Security

HyperShift Isolation

BlueField DPU BlueField DPU BlueField DPU

Infrastructure Cluster

17
 CONFIDENTIAL designator

Thank you linkedin.com/company/red-hat

youtube.com/user/RedHatVideos

Red Hat is the world’s leading provider of

facebook.com/redhatinc
enterprise open source software solutions. Award-
winning support, training, and consulting services
make twitter.com/RedHat
Red Hat a trusted adviser to the Fortune 500.

18