Generative AI and Prompt Engineering with AWS

In this course, students will learn with practical implementation about large language models, prompt engineering techniques, AWS Bedrock foundation models, AWS Bedrock agents, LangChain integration for building conversational chatbots integrated with RAG, and AWS Lambda Functions.

Intermediate 60 Days Weekends
  • Introduction to Large Language Models (LLMs)
    • What are LLMs?
      • Definition and basic concepts
      • How LLMs differ from traditional NLP models
      • The role of deep learning in LLMs
    • History and evolution of LLMs
      • From rule-based systems to statistical models
      • The advent of neural networks in NLP
      • Milestone models: Word2Vec, LSTM, Transformer
      • The GPT family and its impact
  • Project: LLM Exploration and Evaluation
    • Compare different LLMs available on AWS Bedrock, analyzing their strengths and weaknesses for various tasks.
    • Create a report summarizing your findings and present recommendations for specific use cases.
  • Tokenization, Embeddings, Attention Mechanisms
    • Tokenization methods
      • BPE, Wordpiece, SentencePiece
    • Word embeddings vs. Contextual embeddings
    • Self attention and multi-head attention explained
  • AWS Bedrock and Foundation Models
    • Introduction to AWS Bedrock
      • Overview of AWS AI Services ecosystem
      • Bedrock's role in democratizing AI
      • Key features and benefits of Bedrock
    • Available foundation models in Bedrock
      • Overview of models: Claude, Stable Diffusion, Titan, etc.
      • Comparison of model capabilities and use cases
      • Pricing and performance considerations
    • Accessing and using Bedrock APIs
      • Setting up AWS account and permissions
      • API structure and common parameters
      • Hands-on exercise/assignment: Making API calls to Bedrock
  • Project: Custom Chatbot Development
    • Develop a domain-specific chatbot using AWS Bedrock.
    • Select an appropriate foundation model and deploy it as an interactive chatbot.
  • Popular LLM Architectures and Ethical Considerations
    • LLM Architectures
      • GPT (Generative Pre-trained Transformer) architecture
      • BERT (Bidirectional Encoder Representations from Transformers)
      • T5 (Text-to-Text Transfer Transformer)
      • Comparison of architectures and their use cases
    • Limitations and Ethical Considerations
      • Biases in training data and model outputs
      • Hallucinations and factual inconsistencies
      • Privacy concerns and data protection
  • Prompt Engineering Techniques
    • Basics of prompt engineering
      • What is a prompt
      • Components of an effective prompt
      • The art and science of crafting prompts
    • Zero-shot, one-shot, and few-shot learning
      • Zero-shot: Performing tasks without specific example
      • One-shot: Using a single example to guide the model
      • Few-shot: Providing multiple examples (typically 2-5) for improved performance
  • Constrained prompts and Prompt templates and variables
    • Constrained prompts and output formatting
      • Techniques for controlling model output
      • Designing prompts for specific output formats
      • Handling structured data input and output
    • Prompt templates and variables
      • Creating reusable prompt templates
      • managing dynamic content in prompts
      • Best practices for prompt versioning and management

     

  • Zero-shot learning techniques
    • Crafting clear and specific instructions
    • Leveraging the model's pre-trained knowledge
    • Using task-specific keywords and phrases
    • Handling ambiguity and potential misinterpretations
  • One-shot learning strategies
    • Designing effective single examples
    • Balancing specificity and generalizability in the example
    • Techniques for seamlessly transitioning from the example to the target task
  • Few-shot learning best practices
    • Selecting diverse and representative examples
    • Structuring examples for pattern recognition
    • Determining the optimal number of examples for different tasks
    • Techniques for ordering examples to maximize effectiveness
    • Comparative analysis of zero-shot, one-shot, and few-shot approaches
      • Strengths and limitations of each technique
      • Suitable use cases for each technique
      • Performance comparisons across different types of tasks
    • Practical exercises
      • Crafting prompts for each learning type
      • Analyzing model outputs and iterating on prompt design
      • Hands-on practice with AWS Bedrock models
  • Chain-of-thought (CoT) prompting
    • Encouraging step-by-step reasoning
    • Applications in problem-solving and decision-making
    • Combining chain-of-thought with other techniques
  • Project: Multi-task Language Assistant
    • Create a language assistant capable of performing multiple tasks (e.g., translation, summarization, sentiment analysis) using a single LLM.
    • Apply various prompt engineering techniques to optimize performance across different tasks.
  • Context stuffing
    • Techniiques for maximizing context window usage
    • Balancing relevance and quantity of context
    • Handling long documents and large datasets
  • Retrieval-Augmented Generation (RAG)
    • Introduction to RAG
      • Definition of  and core concepts of RAG
      • Advantages of RAG over traditional LLM approaches
    • Practice exercise:
      • Implementing RAG for Question Answering
      • Implementing RAG for document summarization
  • RAG Architecture and Knowledge Bases
    • RAG Pipeline
      • Overview of RAG pipeline
      • Key components: retriever, generator, and knowledge base
    • Types of retrievers:
      • Dense retrievers
      • Sparse retrievers
      • Hybrid retrievers
    • Building Knowledge Bases
      • Data preparation and preprocessing
      • Text chunking strategies
      • Embedding generation and storage
      • Indexing techniques for efficient retrieval
  • Retrieval Mechanisms and Prompt Engineering for RAG
    • Retrieval Mechanisms
      • Dense retrieval using embeddings
      • Sparse retrieval methods (e.g., TF-IDF)
      • Hybrid retrieval approaches
      • Semantic search and similarity metrics
    • Prompt Engineering for RAG
      • Designing effective queries for retrieval
      • Incorporating retrieved context into prompts
      • Balancing retrieved information and task instructions 
    • Implementing RAG with AWS Services
      • Leveraging Amazon OpenSearch for vector search
      • Integrating RAG with AWS Bedrock models
      • Implementing conversational memory in RAG-based chatbots
    • Practice exercise:
      • Developing a conversational chatbot integrated with RAG
  • RAG Project
    • Building a RAG-based question answering system using AWS services
  • AWS Bedrock Agents
    • Introduction to Bedrock Agents
      • Concept of AI agents in Bedrock
      • Comparison with traditional chatbots
      • Use cases and applications
    • Creating and configuring agents
      • Agent creation process in Bedrock
      • Defining agent personality and knowledge base
    • Defining agent behaviors and capabilities
      • Action groups and API integrations
      • Integrating Bedrock agents with AWS Lambda for custom processing
  • LangChain and AWS Integration
    • Introduction to LangChain
      • LangChain's role in LLM application development
      • Key concepts: Chains, Agents, Memory
      • LangChain's modular architecture
    • LangChain components
      • Working with prompt templates
      • Integrating different LLM providers
      • Implementing various memory types
      • Buidling complex chains
  • Integrating LangChain with AWS Services
    • Using Bedrock models in LangChain
    • Deploying LangChain applications on AWS
    • Building complex LLM applications with LangChain
      • Implementing conversational agents
      • Creating document Q&A systems
      • Developing summarization and analysis tools
  • Project: Advanced Question-Answering System with LangChain
    • Build a sophisticated question-answering system using LangChain and AWS services.
    • Incorporate document retrieval, multi-step reasoning, and source citation capabilities.

How students rated this courses

0.0

(Based on 0 reviews)

Reviews

Transcript from the "Introduction" Lesson

Course Overview [00:00:00]

My name is John Deo and I work as human duct tape at Gatsby, that means that I do a lot of different things. Everything from dev roll to writing content to writing code. And I used to work as an architect at IBM. I live in Portland, Oregon.

Introduction [00:00:16]

We'll dive into GraphQL, the fundamentals of GraphQL. We're only gonna use the pieces of it that we need to build in Gatsby. We're not gonna be doing a deep dive into what GraphQL is or the language specifics. We're also gonna get into MDX. MDX is a way to write React components in your markdown.

Why Take This Course? [00:00:37]

We'll dive into GraphQL, the fundamentals of GraphQL. We're only gonna use the pieces of it that we need to build in Gatsby. We're not gonna be doing a deep dive into what GraphQL is or the language specifics. We're also gonna get into MDX. MDX is a way to write React components in your markdown.

A Look at the Demo Application [00:00:54]

We'll dive into GraphQL, the fundamentals of GraphQL. We're only gonna use the pieces of it that we need to build in Gatsby. We're not gonna be doing a deep dive into what GraphQL is or the language specifics. We're also gonna get into MDX. MDX is a way to write React components in your markdown.

We'll dive into GraphQL, the fundamentals of GraphQL. We're only gonna use the pieces of it that we need to build in Gatsby. We're not gonna be doing a deep dive into what GraphQL is or the language specifics. We're also gonna get into MDX. MDX is a way to write React components in your markdown.

Summary [00:01:31]

We'll dive into GraphQL, the fundamentals of GraphQL. We're only gonna use the pieces of it that we need to build in Gatsby. We're not gonna be doing a deep dive into what GraphQL is or the language specifics. We're also gonna get into MDX. MDX is a way to write React components in your markdown.

Course - Frequently Asked Questions

How this course help me to design layout?

My name is Jason Woo and I work as human duct tape at Gatsby, that means that I do a lot of different things. Everything from dev roll to writing content to writing code. And I used to work as an architect at IBM. I live in Portland, Oregon.

What is important of this course?

We'll dive into GraphQL, the fundamentals of GraphQL. We're only gonna use the pieces of it that we need to build in Gatsby. We're not gonna be doing a deep dive into what GraphQL is or the language specifics. We're also gonna get into MDX. MDX is a way to write React components in your markdown.

Why Take This Course?

We'll dive into GraphQL, the fundamentals of GraphQL. We're only gonna use the pieces of it that we need to build in Gatsby. We're not gonna be doing a deep dive into what GraphQL is or the language specifics. We're also gonna get into MDX. MDX is a way to write React components in your markdown.

Is able to create application after this course?

We'll dive into GraphQL, the fundamentals of GraphQL. We're only gonna use the pieces of it that we need to build in Gatsby. We're not gonna be doing a deep dive into what GraphQL is or the language specifics. We're also gonna get into MDX. MDX is a way to write React components in your markdown.

We'll dive into GraphQL, the fundamentals of GraphQL. We're only gonna use the pieces of it that we need to build in Gatsby. We're not gonna be doing a deep dive into what GraphQL is or the language specifics. We're also gonna get into MDX. MDX is a way to write React components in your markdown.

$1,000
Installments
Enroll Now Starts October 18, 2024

What's included

  • Certificate
  • 22 Modules
  • Live Classes
  • Lifetime access
WhatsApp