© 2023, Amazon Web Services, Inc. or its affiliates. All rights reserved.
AIM371
Build and scale generative AI
applications with Amazon Bedrock
Vikesh Pandey Philipp Kaindl
(he/him) (he/him)
Sr. AI/ML Specialist Solutions Architect Sr. AI/ML Specialist Solutions Architect
AWS AWS
© 2023, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Agenda
• Overview of Amazon Bedrock
• Setting up AWS accounts
• Lab 1: Extract insights using prompt engineering techniques
• Lab2: Build marketing email generation application using Amazon
Bedrock
• Q&A, survey, and next steps
© 2023, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Workshop staff
Philipp Kaindl Anastasia Tzeveleka
Vikesh Pandey
Praveen Kumar Jayakumar Aris Tsakpinis
© 2023, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Tingyi Li
Overview of Amazon Bedrock
© 2023, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Amazon Bedrock
THE EASIEST WAY TO BUILD AND SCALE GENERATIVE AI APPLICATION WITH FMS
Access a range of leading FMs via a single API
Privately customize FMs with your own data
Enable data security and compliance
Build agents that execute complex business tasks by dynamically invoking APIs
Extend the power of FMs with your data using Retrieval Augmented Generation (RAG)
No need to manage infrastructure
© 2023, Amazon Web Services, Inc. or its affiliates. All rights reserved.
How Amazon Bedrock works
Amazon Bedrock Choose Use as is or Send Receive
an FM customize prompt response
Build generative AI
applications using FMs Use the playground Fine-tune FMs as Use Amazon Bedrock Receive model
through a serverless to experiment with needed; Amazon API to send your response in your
API service FMs and select Bedrock will prompts to application
the one that suits automatically deploy the model
your needs the FM for inference
© 2023, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Amazon Bedrock supports leading
foundation models
Amazon Titan Jurassic-2 Claude 2
Text summarization, generation, Multilingual LLMs for text generation LLM for conversations, question
classification, open-ended Q&A, in Spanish, French, German, answering, and workflow automation
information extraction, embeddings, Portuguese, Italian, and Dutch based on research into training
and search honest and responsible AI systems
Command and Embed Llama 2 Stable Diffusion
Text generation model for business Fine-tuned models ideal for dialogue Generation of unique, realistic,
applications and embeddings model use cases and language tasks high-quality images, art, logos,
for search, clustering, or and designs
classification in 100+ languages
© 2023, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Amazon Bedrock inference options
On-demand Provisioned throughput
Pay-as-you-go, no commitment Provision sufficient throughput to meet your
application’s performance requirements
› Pricing based on input and output token › Guaranteed throughput at a fixed cost
count for LLMs
› Higher throughput available
› Great for prototyping, POCs, small
workloads with more relaxed requirements › Flexible commitment term of 1 month or
for throughput and latency 6 months
› Requests per minute (RPM) and tokens per › Pay hourly rate, discounted for
minute (TPM) limits enforced extended commitment
› Great for production workloads or
inference on custom models
© 2023, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Data privacy
You are always in control of your data
Customer data is not used to improve Amazon Titan models
for other customers and is not shared with other foundation
model providers
Customer data (prompts, responses, fine-tuned models) remain
in the Region where they are created
© 2023, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Data security
You are always in control of your data
• Support for AWS PrivateLink
• Integration with AWS Identity and Access Management (IAM)
• Integration with AWS CloudTrail
• Encryption support with service-managed keys &
customer-managed keys (CMK)
• VPC support of fine-tuning jobs
© 2023, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Lab instructions
© 2023, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Setup AWS account
• Start a new browser session with no cache (Chrome, Firefox
preferred)
• Open the link: https://bit.ly/aim-371 and use your personal email
to get OTP and sign in
• On next page, Click “I agree with the Terms and Conditions” and
“Join Event“
• You are on Event Dashboard now. On the left pane, click on Open
AWS Console (us-west-2). AWS console will open in a new tab.
• For further instructions, go back to Event Dashboard and follow next
steps
© 2023, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Lab 1: Extract insights using prompt
engineering techniques
Follow instructions in the Lab guide
© 2023, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Lab 2: Build email marketing content
creation application using Amazon
Bedrock
Follow instructions in the Lab guide
© 2023, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Thank you! Please complete the session
survey in the mobile app
Vikesh Pandey Philipp Kaindl
pandvike@amazon.co.uk philikai@amazon.de
© 2023, Amazon Web Services, Inc. or its affiliates. All rights reserved.
