Developing Generative AI Applications on AWS

Course Description

This course is designed to introduce generative artificial intelligence (AI) to software developers interested in using large language models (LLMs) without fine-tuning. The course provides an overview of generative AI, planning a generative AI project, getting started with Amazon Bedrock, the foundations of prompt engineering, and the architecture patterns to build generative AI applications using Amazon Bedrock and LangChain.

Who is this course for

This course is intended for software developers who want to leverage the power of LLMs and foundation models within the AWS ecosystem. It is specifically designed for those who prefer using APIs and orchestration frameworks like LangChain over manual model training or fine-tuning.

Course Objectives

• Fundamental Theory: Describe generative AI alignment with machine learning and identify potential risks and business value.

• Amazon Bedrock Mastery: Understand Bedrock's cost structure, APIs, and architecture patterns; implement console-based demonstrations.

• Prompt Engineering: Apply zero-shot, few-shot, and advanced prompt techniques while mitigating bias and misuse.

• Architectural Patterns: Identify components of a GenAI application and use LangChain to integrate LLMs, chains, and document loaders.

• Practical Application: Build and test real-world use cases using the Retrieval Augmented Generation (RAG) approach and Agents for Amazon Bedrock.

Prerequisites

• Required: Completed AWS Technical Essentials.

• Technical Skill: Intermediate-level proficiency in Python.

Course outline

Section 1: Introduction to Generative AI – Art of the Possible

• Overview of ML

• Basics of generative AI

• Generative AI use cases

• Generative AI in practice

• Risks and benefits

Section 2: Planning a Generative AI Project

• Generative AI fundamentals

• Generative AI in practice

• Generative AI context

• Steps in planning a generative AI project

• Risks and mitigation

Section 3: Getting Started with Amazon Bedrock

• Introduction to Amazon Bedrock

• Architecture and use cases

• How to use Amazon Bedrock

• Demonstration: Setting up Bedrock access and using playgrounds

Section 4: Foundations of Prompt Engineering

• Basics of foundation models

• Fundamentals of prompt engineering

• Basic prompt techniques

• Advanced prompt techniques

• Model-specific prompt techniques

• Demonstration: Fine-tuning a basic text prompt

• Addressing prompt misuses

• Mitigating bias

• Demonstration: Image bias mitigation

Section 5: Amazon Bedrock Application Components

• Overview of generative AI application components

• Foundation models and the FM interface

• Working with datasets and embeddings

• Demonstration: Word embeddings

• Additional application components

• Retrieval Augmented Generation (RAG)

• Model fine-tuning

• Securing generative AI applications

• Generative AI application architecture

Section 6: Amazon Bedrock Foundation Models

• Introduction to Amazon Bedrock foundation models

• Using Amazon Bedrock FMs for inference

• Amazon Bedrock methods

• Data protection and auditability

• Demonstration: Invoke Bedrock model for text generation using zero-shot prompt

Section 7: LangChain

• Optimizing LLM performance

• Using models with LangChain

• Constructing prompts

• Demonstration: Bedrock with LangChain using a prompt that includes context

• Structuring documents with indexes

• Storing and retrieving data with memory

• Using chains to sequence components

• Managing external resources with LangChain agents

Section 8: Architecture Patterns

• Introduction to architecture patterns

• Text summarization

• Demonstration: Text summarization of small files with Anthropic Claude

• Demonstration: Abstractive text summarization with Amazon Titan using LangChain

• Question answering

• Demonstration: Using Amazon Bedrock for question answering

• Chatbot

• Demonstration: Conversational interface – Chatbot with AI21 LLM

• Code generation

• Demonstration: Using Amazon Bedrock models for code generation

• LangChain and agents for Amazon Bedrock

• Demonstration: Integrating Amazon Bedrock models with LangChain agents