0% found this document useful (0 votes)
339 views43 pages

(25D2S01) - (Keynote Speech) AI First

The document outlines the transition from a cloud-first to an AI-first mindset in business, emphasizing the importance of AI strategies in IT purchasing decisions. It discusses the current state of generative AI, highlighting its potential benefits and the challenges organizations face in leveraging it effectively. Additionally, it presents AWS's generative AI stack, showcasing tools and infrastructure designed to enhance productivity and support the development of AI applications.

Uploaded by

이동환
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
339 views43 pages

(25D2S01) - (Keynote Speech) AI First

The document outlines the transition from a cloud-first to an AI-first mindset in business, emphasizing the importance of AI strategies in IT purchasing decisions. It discusses the current state of generative AI, highlighting its potential benefits and the challenges organizations face in leveraging it effectively. Additionally, it presents AWS's generative AI stack, showcasing tools and infrastructure designed to enhance productivity and support the development of AI applications.

Uploaded by

이동환
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 43

DATA&AI ROADSHOW

AI First 시대의 시작

김기완
Director, Solutions Architecture

© 2025, Amazon Web


© 2025,
Services,
Amazon
Inc. or
Webits Services,
affiliates.Inc.
All or
rights
its affiliates.
reserved.All
Amazon
rights Confidential
reserved. Amazon
and Trademark.
Confidential and Trademark. 1
© 2025, Amazon Web Services, Inc. or its affiliates. All rights reserved. Amazon Confidential and Trademark. 2
The Market is Shifting to an AI-First Mindset

Cloud-First AI-First
(Timeframe: previous 10 years) (Timeframe: Now)

Purchasing Legacy cloud purchases Purchasing AI strategy drives IT


Mindset govern future IT Mindset purchasing decisions
purchasing decisions (with little regard to
Key Decision IT Leaders (CIO) previous investments)
Makers Key Decision Business Leaders
Market Cloud market share leaders Makers (IT Leaders advise)
Beneficiaries and their partners Market (Perceived) AI market
(Incumbency wins) Beneficiaries leaders

© 2025, Amazon Web Services, Inc. or its affiliates. All rights reserved. Amazon Confidential and Trademark.
Generative AI Is Not Yet Delivering on the Top
Business Priorities
CEO Top 10 Strategic Business CIO Top Business Value from
Priorities for 2024 to 2025 Applying Generative AI

Revenue Growth
Growth 62% +25% Productivity
Productivity 74%
Customer Experience 49%
Technology-related 33% Digital Transformation 31%
Employee Experience 24%
Workforce 27%
Innovation 18%
Financial 23% Reduce Costs 18%
Business Agility 17%
Corporate 22% Profitability 12%
Cost Management 16% Competitiveness 8%
Business Risk 8%
Customer 10% Time-to-Market 8%
Environmental Sustainability Market Share 5%
10%
Compliance 5%
Products and Services 9% Transform Culture 5%
Supplier Coordination 4%
Efficiency Efficiency
and Productivity
and Productivity 8%
Revenue Growth
Grow revenue 3%
0% 35% 70% 0% 40 % 80 %

What are your organizations’ top three strategic business priorities for the next two years (2024/2025)? What are the top three types of business value your enterprise seeks from applying generative AI?
Source: 2024 Gartner CEO and Senior Business Executive Survey Source: 2025 Gartner CIO and Technology Executive Survey (n=3,067)

© 2025, Amazon Web Services, Inc. or its affiliates. All rights reserved. Amazon Confidential and Trademark. 4
생성형 AI (Generative AI)

© 2025, Amazon Web Services, Inc. or its affiliates. All rights reserved. Amazon Confidential and Trademark. 5
Generative AI: Trends through the years

2023 2024 2025


The year of POCs The year of Production The year of Business Value

What is generative AI? Is this secure? How do I prioritize my projects? How can I leverage agents? How can we go deeper and wider?

Do I need to become a prompt engineer? How can I lower my costs? How do I make this real? How can we transform our entire business?

How do I choose a model? Where do I get started? What customization method should I use? How can I get all my employees to embrace AI?

© 2025, Amazon Web Services, Inc. or its affiliates. All rights reserved.
AWS Generative AI Stack

APPLICATIONS TO BOOST PRODUCTIVITY


Amazon Q Business Amazon Q Developer
INSIGHTS AND AUTOMATION S O F T WA R E D E V E L O P M E N T L I F E C Y C L E

MODELS AND TOOLS TO BUILD GENERATIVE AI APPS


Amazon Bedrock
AMAZON MODELS | PARTNER MODELS

INFRASTRUCTURE TO BUILD AND TRAIN AI MODELS


Amazon SageMaker AI AWS Trainium
GPUs
MANAGED INFRASTRUCTURE AWS Inferentia
HIGH PERFORMANCE COMPUTE

© 2025, Amazon Web Services, Inc. or its affiliates. All rights reserved.
7
AWS Generative AI Stack

APPLICATIONS TO BOOST PRODUCTIVITY


Amazon Q Business Amazon Q Developer
INSIGHTS AND AUTOMATION S O F T WA R E D E V E L O P M E N T L I F E C Y C L E

MODELS AND TOOLS TO BUILD GENERATIVE AI APPS


Amazon Bedrock
AMAZON MODELS | PARTNER MODELS

INFRASTRUCTURE TO BUILD AND TRAIN AI MODELS


Amazon SageMaker AI AWS Trainium
GPUs
MANAGED INFRASTRUCTURE AWS Inferentia
HIGH PERFORMANCE COMPUTE

© 2025, Amazon Web Services, Inc. or its affiliates. All rights reserved.
8
Build FM’s from scratch
Create your own ML models, including FMs, with integrated
purpose-built tools and high-performance, cost-effective
infrastructure

Customize foundation models


Amazon Access and evaluate 250+ FMs that can be customized
easily for your use case

SageMaker AI Implement MLOps & governance


Create reliable and repeatable workflows incorporating
Build, train, and deploy ML MLOps practices with purpose-built tooling
models at scale, including FMs

Improve ML governance
Enhance model governance and compliance with built-in
governance tools

Manage and deploy models for inference


Easiest way to deploy AI & ML models including foundation models
(FMs) to make inference requests at the best price performance for
any use case
© 2025, Amazon Web Services, Inc. or its affiliates. All rights reserved.
9
Innovating at the silicon level

AWS AWS
AWS AI Chips

© 2025, Amazon Web Services, Inc. or its affiliates. All rights reserved.
10
CUSTOMER HIGHLIGHT

models run 60% faster with Trainium2

At Anthropic, our mission is to ensure transformative AI


helps people and society flourish. Speed is essential for
our enterprise users, that’s why we announced the
availability of Claude Haiku 3.5, which matches the
performance of Claude 3 Opus, at industry leading
latency,” said Dario Amodei. “We are excited to
introduce Claude 3.5 Haiku running on Trn2 for
latency-optimized inference, now available via Amazon
Bedrock. It provides the fastest response time for our
customers, 60% faster than any other instance we use.
We are excited to continue migrating more Claude
models to Trainium2, so our customers enjoy the
performance boost and productivity increase.”
Dario Amodei
© 2025, Amazon Web Services, Inc. or its affiliates. All rights reserved.
11
Founder| Anthropic
CUSTOMER HIGHLIGHT

reduces cost by 40% with AWS Trainium2

“At Poolside, we're building foundation models, an API,


and an Assistant to bring the power of generative AI to
your developers' hands (or keyboard). A major key to
enable this technology, is the infrastructure we are
using to build and run our products. With AWS
Trainium2 our customers will be able to scale their
usage of poolside at a price performance ratio unlike
other AI accelerators. In addition, we plan to train
future models with Trainium2 UltraServers with
expected savings of 40% compared to EC2 P5
instances.”

Eiso Kant

CTP & Co-Founder | Poolside


© 2025, Amazon Web Services, Inc. or its affiliates. All rights reserved.
12
AWS

Third-generation chip purpose built


for Generative AI and ML training
Dense Compute Sparse Compute

HBM Capacity HBM Bandwidth

© 2025, Amazon Web Services, Inc. or its affiliates. All rights reserved.
13
Amazon EC2 Trn2 UltraServer
2 racks, 4 servers, connected together via NeuronLink-v3
providing the highest performance ML compute in EC2

DEN SE C O M PU TE SPARSE C O M PU TE HBM BAN DW IDTH

HBM C APAC ITY N EURO N LIN K EFAv3 BAN DW IDTH


BAN DW IDTH

© 2025, Amazon Web Services, Inc. or its affiliates. All rights reserved.
14
Project Rainier

Hundreds of thousands of AWS


Trainium2 chips with with the most
exaFLOPs in a single training cluster

more compute than


Anthropic’s next biggest
training cluster
© 2025, Amazon Web Services, Inc. or its affiliates. All rights reserved.
15
© 2024, Amazon Web Services, Inc. or its affiliates. All rights reserved.
AWS Generative AI Stack

APPLICATIONS TO BOOST PRODUCTIVITY


Amazon Q Business Amazon Q Developer
INSIGHTS AND AUTOMATION S O F T WA R E D E V E L O P M E N T L I F E C Y C L E

MODELS AND TOOLS TO BUILD GENERATIVE AI APPS


Amazon Bedrock
AMAZON MODELS | PARTNER MODELS

INFRASTRUCTURE TO BUILD AND TRAIN AI MODELS


Amazon SageMaker AI AWS Trainium
GPUs
MANAGED INFRASTRUCTURE AWS Inferentia
HIGH PERFORMANCE COMPUTE

© 2025, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Amazon Choice of leading FMs through a single API

Bedrock Model customization

The easiest way to build and


Retrieval Augmented Generation (RAG)
scale generative AI applications
with foundation models (FMs)
Agents that execute multistep tasks

Security, privacy, and safety

© 2025, Amazon Web Services, Inc. or its affiliates. All rights reserved.
None of the customer’s data is used to
train the underlying model

All data is encrypted in transit and at rest;

Amazon data used for customization is securely


transferred through customer’s VPC

Helps keep your data


Data remains in the Region where the API
secure and private is processed

Support for GDPR, SOC, ISO, CSA


compliance, and HIPAA eligibility

© 2025, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Amazon Bedrock
BROAD CHOICE OF MODELS

Effective Frontier Advanced Multimodal search High-quality Advanced Knowledge Software High-quality AI
reasoning & rapid multimodal reasoning & & advanced video generation image & summarization, engineering AI image
analysis for long intelligence at low- coding retrieval powering from text & language expert agents, & for large generation,
context windows latency, Agent & capabilities, multilingual images reasoning code completion enterprises easily deployable
RAG Applications, including knowledge agents at scale
high-quality image computer use
& video generation skills

JAMBA AMAZON NOVA CLAUDE COMMAND LUMA RAY 2 LLAMA MISTRAL MALIBU STABLE DIFFUSION

EMBED MIXTRAL POINT STABLE IMAGE

RERANK

Coming soon

Amazon Bedrock Marketplace enables developers to discover, test, and use over 100 popular, emerging, and specialized foundation models (FMs) alongside
the current selection of industry-leading models in Amazon Bedrock. DeepSeek-R1 model is now available in Amazon Bedrock Marketplace.

© 2025, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Amazon Nova Foundation Models

State-of-the-art foundation models that deliver frontier intelligence


and industry leading price performance.

Understanding models Creative content generation models

Amazon Nova Amazon Nova Amazon Nova Amazon Nova Amazon Nova Amazon Nova
Micro Lite Pro Premier Canvas Reel
Our most capable State-of-the-art State-of-the-art
Our text only Our lowest cost Our highly capable
multimodal model multimodal model multimodal model image generation video generation
model that for complex
that is lightning with best model model
delivers the lowest reasoning tasks
latency responses fast for combination of
lightweight tasks accuracy, speed, and and for use as the
at very low cost best teacher for
cost for a wide range
of tasks distilling custom
models
GENERALLY AVAILABLE GENERALLY AVAILABLE GENERALLY AVAILABLE GENERALLY AVAILABLE
GENERALLY AVAILABLE COMING SOON

Lower Cost Increasing


& Latency Intelligence

© 2025, Amazon Web Services, Inc. or its affiliates. All rights reserved.
21
AWS Generative AI Stack

APPLICATIONS TO BOOST PRODUCTIVITY


Amazon Q Business Amazon Q Developer
INSIGHTS AND AUTOMATION S O F T WA R E D E V E L O P M E N T L I F E C Y C L E

MODELS AND TOOLS TO BUILD GENERATIVE AI APPS


Amazon Bedrock
AMAZON MODELS | PARTNER MODELS

INFRASTRUCTURE TO BUILD AND TRAIN AI MODELS


Amazon SageMaker AI AWS Trainium
GPUs
MANAGED INFRASTRUCTURE AWS Inferentia
HIGH PERFORMANCE COMPUTE

© 2025, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Unites your data sources, repositories, and
enterprise systems

Amazon Q
Available where you work
Reinvent work with AWS’
generative AI–powered assistant
Delivers industry-leading assistance

© 2025, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Reinvent work with
BUSINESS DEVELOPERS SPECIALIZED USERS

Amazon Q Amazon Q in Amazon Q Amazon Q in


Business QuickSight Developer Connect

KNOWLEDGE SEARCH UNDERSTAND DATA PLAN APPLICATION AGENT ASSIST

SUMMARIZATION BUILD & REFINE VISUALS CODE GENERATION

CONTENT CREATION BUILD CALCULATIONS UNIT TESTING


Amazon Q in
EXTRACT INSIGHTS EXECUTIVE SUMMARIES SECURITY SCANNING AWS Supply Chain

RESEARCH & ANALYSIS CREATE DATA STORIES CODE REMEDIATION SUPPLY CHAIN

CODE MIGRATION

TROUBLESHOOTING

DEVELOPER KNOWLEDGE
© 2025, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Ingest document
content and
permissions
information
Built for the
enterprise Content
Query
User and group info

Permissions
Identity Permissions User
filtered response
Provider
Amazon Q
Business

© 2025, Amazon Web Services, Inc. or its affiliates. All rights reserved.
< one hour coding time
The median developer spends less than one hour a day writing code1

Developer time spent Developer time


on running and on innovation and
maintaining applications transformation

73% 27%
© 2025, Amazon Web Services, Inc. or its affiliates. All rights reserved. Amazon Confidential and Trademark. 1 Software.com, “Global Code Time Report,” January 2022
Helps developers and IT professionals build faster across
the entire software development lifecycle (SDLC)
Amazon Q Most accurate coding recommendations
Developer
Agents can autonomously help you implement features,
refactor code, perform software upgrades and more

Amazon Q is an expert on AWS and optimizing AWS


environment

Best-in-class security vulnerability scanning and


remediation

Amazon Q is built with security and privacy in mind from the start, making it
easier for organizations to use generative AI safely.
© 2025, Amazon Web Services, Inc. or its affiliates. All rights reserved. Amazon Confidential and Trademark.
Maintain, Modernize, and Transform
AMAZON Q CODE TRANSFORMATION

.NET Framework Analysis Planning


Cross-platform .NET Up to
applications Refactoring
applications on Linux
on Windows Compatibility tests 4x
faster
Documentation Wave
COBOL applications Java applications
planning Decomposition
on mainframes on AWS Up to
Refactoring
40%
savings
Discovery Wave planning
VMware workloads Network automation
VMware workloads
on-premises on Amazon EC2
Workload migration Access to
a range
Java applications
Build/test v8/v11 Apply
Java applications of LLMs
knowledge base Build/test
in v8, v11 v17Error fixing using genAI
in v17 and v21

© 2025, Amazon Web Services, Inc. or its affiliates. All rights reserved. Amazon Confidential and Trademark.
Data

© 2025, Amazon Web Services, Inc. or its affiliates. All rights reserved. Amazon Confidential and Trademark.
Generative AI
Application

Data
Foundation
STORAGE

GOVERNANCE
& COMPLIANCE

DATABASES,
ANALYTICS,
& DATA LAKES

DATA
INTEGRATION

© 2024, Amazon Web Services, Inc. or its affiliates. All rights reserved. 31
Convergence of analytics and AI jobs to be done,
often using the same data

Convergence of roles across data analyst,


data scientist, and data engineer

Data silos drive up costs and complexity

Fragmented governance erodes trust and


confidence

© 2024, Amazon Web Services, Inc. or its affiliates. All rights reserved.
NEW

The next generation of


Amazon SageMaker
The center for all your data, analytics, and AI

© 2024, Amazon Web Services, Inc. or its affiliates. All rights reserved.
a single data and AI development environment
NEW

The next generation of

Amazon broadest set of tools for AI use cases

SageMaker
an open lakehouse to unify all your data

built-in data and AI governance

© 2024, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Amazon SageMaker

Unified Studio (preview)

Data SQL Model Gen AI App Streaming Search Business


Processing Analytics Development Development Intelligence
coming soon coming soon
coming soon

Data & AI Governance

Lakehouse

© 2024, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Amazon SageMaker

Amazon SageMaker Unified Studio


(preview)
Data Processing SQL Analytics Model Development Gen AI APP Streaming Search Business Intelligence
Development coming soon coming soon coming soon

© 2024, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Data & AI Governance
Seamless access to all data, all resources, and all tools

DATA CODE MODELS COMPUTE

JUPYTER NOTEBOOKS SQL EDITOR BEDROCK IDE


(preview)

© 2024, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Collaborate and build faster

Train and deploy AI models Build custom Gen AI apps


with Amazon SageMaker AI with Bedrock IDE (preview)

Prepare and integrate Run SQL queries with


data with Amazon EMR Amazon Redshift

© 2024, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Accelerated by Amazon Q Developer …

…the most capable generative AI assistant for


software development

© 2024, Amazon Web Services, Inc. or its affiliates. All rights reserved.
SageMaker Lakehouse

GENERALLY AVAILABLE

© 2024, Amazon Web Services, Inc. or its affiliates. All rights reserved.
SageMaker
Lakehouse
Unifying data across data warehouses, data lakes,
operational databases, and applications with a single
copy of data

Zero-ETL, federation, sharing

AWS DATA SOURCES | THIRD PARTY DATA SOURCES

© 2024, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Amazon SageMaker Third Party
Unified Studio applications

Fine grained access controls

SageMaker
Lakehouse

Zero-ETL, federation, sharing

AWS DATA SOURCES | THIRD PARTY DATA SOURCES

© 2024, Amazon Web Services, Inc. or its affiliates. All rights reserved.
SageMaker Data & AI
Governance
GENERALLY AVAILABLE

© 2024, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Amazon SageMaker Data and AI Governance
DISCOVER, GOVERN, AND COLLABORATE ON DATA AND AI SECURELY

Unified Studio

Data Models GenAI Compute

SageMaker Catalog Built on Amazon


DataZone, Amazon Q
integrated
Data & AI Governance

Data Permissions and Auditable –


Data Quality Data & ML Guardrails Responsible Cost logging
Classification Lineage Subscriptions AI
& Monitoring

© 2024, Amazon Web Services, Inc. or its affiliates. All rights reserved. 44
Q&A

© 2025, Amazon Web Services, Inc. or its affiliates. All rights reserved. Amazon Confidential and Trademark. 45

You might also like