Curriculum Overview: AI Concepts and Terminology

This curriculum is designed to prepare learners for the AWS Certified AI Practitioner (AIF-C01) exam. It covers the foundational pillars of Artificial Intelligence, Machine Learning, and Generative AI, specifically focusing on how these technologies are implemented within the AWS ecosystem.

Prerequisites

Before starting this curriculum, learners should have a basic understanding of the following:

• Basic Cloud Literacy: Familiarity with cloud computing concepts (e.g., storage, compute, and APIs).
• Data Fundamentals: An understanding of what data is (structured vs. unstructured) and how it is used in a business context.
• Business Logic: Ability to identify business problems that might benefit from automation or prediction.
• No Coding Required: While technical, this curriculum focuses on high-level concepts and managed services rather than deep programming or advanced calculus.

Module Breakdown

The curriculum is structured into six core units, progressing from theoretical foundations to practical AWS implementations.

Learning Objectives per Module

Unit 1: Fundamentals of AI and Machine Learning

• The Hierarchy: Distinguish between Artificial Intelligence (broadest), Machine Learning (subset), and Deep Learning (specialized subset using neural networks).
• Data Savvy: Identify different data formats such as Structured (tabular data like Excel), Unstructured (images/video), and Time-Series (stock prices over time).
• Inferencing: Explain the difference between Batch Inference (processing bulk data at intervals) and Real-time Inference (immediate response to a user request).

Unit 2 & 3: Generative AI and Foundation Models

• Mechanics: Define Tokens (the basic units of text processed by models) and Embeddings (numerical vector representations of data).
• Prompt Engineering: Apply techniques like Few-shot prompting (providing examples in the prompt) to improve model accuracy.
• Adaptation: Compare RAG (Retrieval-Augmented Generation) vs. Fine-tuning for providing models with specialized knowledge.

Unit 4 & 5: Responsible AI and Security

• Fairness & Bias: Identify types of bias in datasets and how they affect model outcomes (e.g., demographic subgroup inaccuracy).
• Explainability: Use tools like Amazon SageMaker Model Cards to document model intentions and limitations.
• Security: Implement the AWS Shared Responsibility Model to secure AI systems, protecting against threats like Prompt Injection.

Success Metrics

To demonstrate mastery of this curriculum, the learner must be able to:

1. Categorize Use Cases: Correctly identify if a problem requires a Classification (Labeling), Regression (Predicting a number), or Clustering (Finding patterns) approach.
2. Service Selection: Choose the correct AWS service for a task (e.g., use Amazon Rekognition for image analysis or Amazon Bedrock for accessing LLMs).
3. Lifecycle Mapping: Describe the steps of the ML Development Lifecycle using the correct terminology.

Real-World Application

Understanding AI concepts and AWS terminology is not just for passing an exam; it has direct career applications:

• Predictive Maintenance: Using time-series data from factory sensors to predict when a machine will fail before it happens.
• Customer Experience: Implementing Amazon Lex to build conversational chatbots that handle 80% of routine customer queries.
• Fraud Detection: Utilizing ML models to analyze transaction patterns in real-time to flag suspicious activity.
• Responsible Innovation: Ensuring that a company's AI tools do not accidentally discriminate against specific groups by monitoring for bias using SageMaker Clarify.

[!IMPORTANT] AI is a rapidly evolving field. This curriculum focuses on the "Foundational" layer, meaning it prioritizes understanding the what and the why over the specialized implementation of the how.

Hands-On Lab: Exploring AI Concepts with AWS Managed Services

Welcome to this guided lab! In this session, we will bridge the gap between theoretical Artificial Intelligence (AI) concepts and practical implementation. You will work with unstructured data, perform real-time inferencing, and see the difference between Computer Vision (CV) and Natural Language Processing (NLP) using AWS managed services.

Prerequisites

Before you begin, ensure you have the following ready:

• AWS Account: An active AWS account with an IAM user or role that has AdministratorAccess (or permissions for Amazon S3, Amazon Rekognition, and Amazon Comprehend).
• AWS CLI: The AWS Command Line Interface installed and configured (aws configure) with your credentials and a default region (e.g., us-east-1).
• Basic Terminal Knowledge: Familiarity with running commands in a bash/zsh or PowerShell terminal.
• Conceptual Understanding: Basic awareness of AI, ML, Computer Vision, and NLP as covered in the study guide.

Concept Review: The AI Hierarchy

Before we build our architecture, let's review the relationship between AI, Machine Learning (ML), and Deep Learning (DL). The services we will use today abstract away the deep learning layers, allowing you to focus on the AI application.

Learning Objectives

By the end of this lab, you will be able to:

1. Differentiate Data Types: Handle unstructured data (images and raw text) in an AI pipeline.
2. Execute Real-Time Inferencing: Send single-request data to pre-trained foundation models and receive immediate predictions.
3. Apply Computer Vision (CV): Use Amazon Rekognition to extract Optical Character Recognition (OCR) data from an image.
4. Apply Natural Language Processing (NLP): Use Amazon Comprehend to determine the sentiment of a text string.

Architecture Overview

The following flowchart visualizes the data flow for this lab. We will use the CLI to interact with unstructured data (an image), extract text using a CV model, and then analyze text using an NLP model.

Step-by-Step Instructions

Step 1: Create a Storage Bucket for Unstructured Data

AI models require data. In this step, we will create an Amazon S3 bucket to hold our unstructured data (images).

[!TIP] S3 bucket names must be globally unique. Replace <YOUR_ACCOUNT_ID> with your actual AWS account number or a random string of numbers to ensure the name is available.

▶Console alternative

1. Navigate to the S3 Console.
2. Click Create bucket.
3. Enter brainybee-lab-ai-concepts-<YOUR_ACCOUNT_ID> as the Bucket name.
4. Leave all other settings as default and click Create bucket.

📸 Screenshot: S3 Create Bucket Form

Step 2: Download and Upload an Image

We need an image containing text to test our Computer Vision model. We will download a sample image and upload it to our new S3 bucket.

▶Console alternative

1. Download the image from this link to your computer.
2. Navigate to your newly created S3 bucket in the console.
3. Click Upload, select your downloaded sample-text.jpg, and click Upload.

Checkpoints

Let's verify our unstructured data is safely stored in S3.

Expected Output: You should see sample-text.jpg listed with its timestamp and file size.

Step 3: Perform Real-Time Inferencing with Computer Vision (CV)

Now, we will use Amazon Rekognition, a managed Deep Learning service for Computer Vision. We will perform real-time inferencing to detect text (OCR) within our unstructured image.

[!NOTE] Look at the JSON output. This represents a model prediction. The service returns DetectedText and a Confidence score (a percentage indicating how certain the ML algorithm is about its prediction).

▶Console alternative

1. Navigate to the Amazon Rekognition Console.
2. In the left sidebar, choose Text in image.
3. Expand the Upload dropdown and upload your local sample-text.jpg.
4. View the extracted text results and confidence scores in the Results pane on the right.

📸 Screenshot: Rekognition Text Detection Results

Step 4: Perform Real-Time Inferencing with Natural Language Processing (NLP)

Next, let's explore Natural Language Processing (NLP). We will use Amazon Comprehend to analyze the sentiment of a sentence. This is an example of taking unstructured text and turning it into structured, labeled insights.

[!TIP] The output will show a Sentiment (e.g., POSITIVE) and a SentimentScore breakdown for Positive, Negative, Neutral, and Mixed. This highlights how AI models deal with probability rather than absolute certainty.

▶Console alternative

1. Navigate to the Amazon Comprehend Console.
2. In the left sidebar, click Real-time analysis.
3. Under Input text, paste: Artificial Intelligence is transforming the world in amazing ways!
4. Click Analyze.
5. Scroll down to the Insights tab and view the Sentiment results.

Teardown

[!WARNING] Remember to run the teardown commands to avoid ongoing charges. Leaving data in S3 incurs minor storage costs.

To clean up your AWS environment, empty and delete the S3 bucket, and remove your local file.

Troubleshooting

Cost Estimate

• Amazon S3: First 50 TB/month is fractions of a cent per GB. Uploading one small image falls well within the AWS Free Tier.
• Amazon Rekognition: Free tier allows 1,000 images per month. Outside free tier, ~$0.001 per image.
• Amazon Comprehend: Free tier allows 50,000 units of text (100 characters per unit) per month. Outside free tier, ~$0.0001 per unit.
• Total Expected Cost: $0.00 (Assuming free tier eligibility or just a few cents otherwise).

Hands-On Lab: Exploring Basic AI Concepts and Terminology

Welcome to this guided hands-on lab! This module bridges the theoretical concepts from Chapter 1 (Fundamentals of AI and ML) with practical reality. You will interact with pre-trained Artificial Intelligence (AI) models using AWS managed services to perform Computer Vision and Natural Language Processing (NLP) tasks.

By completing this lab, abstract terms like unstructured data, real-time inferencing, and confidence scores will become tangible.

Prerequisites

Before starting, ensure you have the following:

• AWS Account: An active AWS account with administrative or developer access.
• CLI Tools: AWS CLI installed and configured (aws configure) on your local machine or AWS CloudShell.
• Prior Knowledge: Familiarity with basic terminal commands and a conceptual understanding of AI vs. Machine Learning (ML).

Learning Objectives

1. Differentiate AI fields in practice: Use distinct services for Computer Vision (Amazon Rekognition) and Natural Language Processing (Amazon Comprehend).
2. Understand Data Types: Work with unstructured data formats (images and raw text) rather than tabular data.
3. Perform Real-Time Inferencing: Send data to an ML model and receive immediate probabilistic predictions.
4. Analyze Model Outputs: Interpret JSON responses containing confidence scores to understand the probabilistic nature of AI.

Architecture Overview

This lab uses a simple serverless architecture to demonstrate real-time inferencing on unstructured data.

Step-by-Step Instructions

Step 1: Prepare Unstructured Data (Amazon S3)

In Machine Learning, images and text are considered unstructured data. Deep learning models are particularly adept at processing this format. Let's create a storage bucket and upload a sample image.

📸 Screenshot: A terminal showing the successful upload: sample.jpg to s3://... output.

▶Console alternative

1. Navigate to Amazon S3 in the AWS Console.
2. Click Create bucket, name it brainybee-lab-vision-<YOUR_ACCOUNT_ID>, and click Create bucket.
3. Select your new bucket, click Upload, and upload a local image file named sample.jpg.

Step 2: Real-Time Inferencing with Computer Vision

Computer Vision is an AI field enabling systems to derive information from images. We will use Amazon Rekognition to perform real-time inferencing (getting immediate predictions) on our unstructured image.

[!TIP] Look at the JSON response. Notice the Confidence attribute next to each label. AI models do not return absolute facts; they return probabilities based on their training data.

▶Console alternative

1. Navigate to Amazon Rekognition in the AWS Console.
2. In the left sidebar, click Label detection.
3. Under the demo section, expand the Upload panel and upload your sample.jpg.
4. Observe the detected labels and their percentage scores (confidence) in the Results pane.

Step 3: Natural Language Processing (NLP)

Natural Language Processing (NLP) allows computers to interpret human language. We will use Amazon Comprehend to analyze the sentiment of unstructured text.

📸 Screenshot: The JSON output from Comprehend showing a Sentiment of MIXED or POSITIVE, along with SentimentScore breakdowns.

▶Console alternative

1. Navigate to Amazon Comprehend in the AWS Console.
2. Scroll down to the Real-time analysis section.
3. Paste the text: "Artificial intelligence is a fascinating and transformative technology, though data drift can occasionally cause headaches!"
4. Click Analyze and view the Sentiment results in the Insights tab.

Checkpoints

Verify your progress after completing the steps above:

• Checkpoint 1: Run aws s3 ls s3://brainybee-lab-vision-<YOUR_ACCOUNT_ID>. You should see sample.jpg listed.
• Checkpoint 2: In your Rekognition JSON output, did you receive a list of Labels? If you used a city image, you should see labels like City, Building, or Urban.
• Checkpoint 3: In your Comprehend JSON output, locate the SentimentScore object. Notice how the scores for Positive, Negative, Neutral, and Mixed all add up to approximately 1.0 (100%).

Concept Review

To solidify your understanding of Chapter 1, let's visually review the hierarchy of AI and the services you just used.

Terminology Mapping Table

Clean-Up / Teardown

[!WARNING] Remember to run the teardown commands to avoid ongoing charges. While Rekognition and Comprehend charge per-request, S3 storage incurs ongoing costs.

Execute the following commands to delete the resources created in this lab:

Note: Amazon Rekognition and Amazon Comprehend APIs used in this lab do not provision permanent infrastructure; they are stateless API calls.

Troubleshooting

Stretch Challenge

Want to test your skills further without step-by-step guidance?

Challenge: Try passing an image containing printed text (like a picture of a street sign or a book page) to Amazon Rekognition to extract the words. Hint: Look up the detect-text API method for Rekognition.

▶Show solution

bashCopy

# Upload an image with text first aws s3 cp sign.jpg s3://brainybee-lab-vision-<YOUR_ACCOUNT_ID>/ # Run the detect-text API (Optical Character Recognition - OCR) aws rekognition detect-text \ --image '{"S3Object":{"Bucket":"brainybee-lab-vision-<YOUR_ACCOUNT_ID>","Name":"sign.jpg"}}'

Curriculum Overview: Mastering Amazon Bedrock and Amazon Q

This curriculum provides a comprehensive roadmap for understanding and implementing AWS's primary generative AI platforms. Learners will move from foundational model access via Amazon Bedrock to specialized AI assistance with Amazon Q.

Prerequisites

Before starting this curriculum, students should possess:

• Cloud Fundamentals: Basic knowledge of AWS infrastructure (IAM, VPCs, and S3).
• AI Literacy: Understanding of basic GenAI terms (Tokens, LLMs, and Hallucinations).
• Security Basics: Familiarity with the AWS Shared Responsibility Model.

---

Module Breakdown

---

Module Objectives per Module

Module 1: Amazon Bedrock Foundations

• Model Access: Explain how to access multiple Foundation Models (FMs) through a single API.
• Parameter Tuning: Master the effects of $Temperature$ and $Top P$ on model creativity.
• Playgrounds: Utilize text, image, and video playgrounds for rapid prototyping.

Module 2: Building with Bedrock

• Retrieval-Augmented Generation (RAG): Implement Knowledge Bases for proprietary data.
• Agentic AI: Configure Amazon Bedrock Agents to execute multi-step business tasks.
• Model Evaluation: Use ROUGE and BLEU scores to assess model performance.

Module 3: Amazon Q Business

• Unified Search: Index corporate data across Slack, Microsoft 365, and SharePoint.
• Q Apps: Create no-code applications for content generation and workflow automation.
• Transparency: Utilize citations and references to ensure response accuracy.

Module 4: Amazon Q Developer

• Development Speed: Increase coding velocity by up to 80% using IDE plugins.
• Modernization: Use agents for heavy-duty tasks like Java or .NET migrations.
• AWS Integration: Query AWS account resources and billing directly via the console.

---

Success Metrics

How to know you have mastered the curriculum:

1. Deployment Proficiency: Successfully deploy an Amazon Bedrock Agent that triggers a Lambda function.
2. Productivity Gains: Demonstrate a measurable reduction in coding time using Amazon Q Developer.
3. Accuracy Verification: Maintain a hallucination rate below a defined threshold using Negative Prompts and Guardrails.
4. Financial Efficiency: Optimize model selection based on token pricing and performance needs.

[!IMPORTANT] Success is not just building a model; it is building a safe model. Ensure all applications utilize Amazon Bedrock Guardrails to prevent prompt injection and data leakage.

---

Real-World Application

Why this curriculum matters in a career:

• Legacy Modernization: Amazon used Q Developer to migrate tens of thousands of applications to Java 17, saving $260 million and 4,500 years of manual labor.
• Employee Efficiency: Amazon Q Business acts as a 24/7 subject matter expert, reducing the time employees spend searching for internal documentation.
• Lower Entry Barrier: Platforms like PartyRock allow non-developers to create AI tools, democratizing innovation across business units.

Comparison: Bedrock vs. Q

▶Click to expand: Specific Business Use Cases

• Automated Inventory: Using Bedrock to monitor supply chain data and recommend reorder points.
• Code Debugging: Using Q Developer to identify security vulnerabilities in Python or Java code.
• Content Creation: Using Q Apps to generate marketing emails based on internal product manuals.

Hands-On Lab: Building and Automating with Amazon Bedrock and Amazon Q

Welcome to this guided lab. In this session, you will explore the powerful synergy between Amazon Bedrock (a managed service for Foundation Models) and Amazon Q Developer (a generative AI-powered assistant). You will use Amazon Q to help generate commands and solve coding challenges while directly interacting with Bedrock to test model inference parameters like Temperature and Top P.

---

Prerequisites

Before starting this lab, ensure you have the following:

• AWS Account: Active account with Administrator or PowerUser access.
• CLI Tools: AWS CLI v2 installed and configured (aws configure).
• IAM Permissions: Policies allowing bedrock:InvokeModel, bedrock:ListFoundationModels, and access to Amazon Q Developer.
• Prior Knowledge: Basic familiarity with terminal commands and JSON structures.

[!IMPORTANT] Amazon Bedrock models are not enabled by default. You must request access to the models in your AWS region before invoking them.

---

Learning Objectives

By completing this lab, you will be able to:

1. Navigate and use Amazon Q Developer to generate accurate AWS CLI commands.
2. Enable and query Foundation Models (FMs) using Amazon Bedrock.
3. Understand and manipulate inference parameters (Temperature, Top P) to control model determinism.
4. Analyze the JSON request and response payloads of Bedrock's InvokeModel API.

---

Architecture Overview

The following diagram illustrates the flow of our lab. We will use Amazon Q as our intelligent assistant to formulate the correct API calls, which we will then send to Amazon Bedrock to interact with the Titan Foundation Model.

---

Step-by-Step Instructions

Step 1: Request Model Access

Before using Amazon Bedrock, you must request access to the specific Foundation Models you intend to use. For this lab, we will use Amazon Titan Text G1 - Lite.

▶Console alternative (Required for first-time setup)

1. Open the AWS Management Console and navigate to Amazon Bedrock.
2. In the left navigation pane, scroll down to Model access.
3. Click Enable specific models (or Manage model access).
4. Check the box next to Titan Text G1 - Lite under the Amazon provider.
5. Click Request model access and wait for the status to change to "Access granted".

📸 Screenshot: Look for the green "Access granted" badge next to the model name.

[!TIP] Some models (like Anthropic Claude) require submitting an additional use-case justification form. Amazon Titan models are usually granted instantly.

Step 2: Use Amazon Q Developer to Generate Commands

Instead of memorizing complex CLI syntax, let's ask Amazon Q to help us figure out how to invoke the Bedrock model.

If you have the Amazon Q CLI integration installed in your terminal, or if you are using the Amazon Q chat pane in the AWS Console, type the following prompt:

"Write an AWS CLI command to invoke the amazon.titan-text-lite-v1 model in Amazon Bedrock to explain what Generative AI is. Output the response to a file called response.txt."

Amazon Q should provide an explanation and a command similar to the one we will use in the next step.

Step 3: Invoke the Foundation Model

Now, let's execute the command to query the Foundation Model. We will pass a JSON payload containing our prompt and configuration.

▶Console alternative

1. In the Bedrock console, go to Playgrounds > Text.
2. Select Amazon as the category and Titan Text G1 - Lite as the model.
3. Type "Explain generative AI in two sentences." in the prompt area.
4. Click Run to see the output.

Step 4: Inspect the Output and Adjust Parameters

The output is saved to response.json. Let's inspect it.

[!NOTE] The inference parameters we used were temperature: 0.1 and topP: 0.9. A low temperature (closer to 0) makes the model more deterministic and focused.

Let's experiment by increasing the temperature to make the model more creative (and potentially less predictable).

Review the new output using cat creative_response.json.

---

Checkpoints

Verify that you have successfully completed the core lab steps:

1. Check 1: Run aws bedrock list-foundation-models --region <YOUR_REGION> | grep titan-text-lite-v1. Does it return the model ID?
2. Check 2: Inspect the response.json file. It should contain a structured JSON response with a results array and an outputText field containing the generated text.

---

Visualizing the Parameter Effect

Below is a conceptual diagram illustrating how parameters like Temperature and Top-P influence the model's token selection process during inference.

---

Teardown

Because this lab primarily uses serverless, on-demand inference, there are no long-running EC2 instances or provisioned endpoints to delete. However, to keep your environment clean, remove the local files generated.

[!WARNING] If you configured Provisioned Throughput for Amazon Bedrock (not covered in this lab but possible in production), you must delete it via the console or CLI, as it incurs high hourly charges.

---

Troubleshooting

---

Stretch Challenge

Now that you understand the CLI interactions, try automating this process with code!

Goal: Write a Python script using the boto3 library that accepts a user string, sends it to the same Titan model, and prints only the outputText string to the console (omitting the rest of the JSON wrapper).

Constraint: You must use Amazon Q Developer (either in your IDE or the console) to help you write the code.

▶Click here to reveal a solution

pythonCopy

import boto3 import json def invoke_titan(prompt): client = boto3.client('bedrock-runtime', region_name='us-east-1') payload = { "inputText": prompt, "textGenerationConfig": { "temperature": 0.7, "topP": 0.9 } } response = client.invoke_model( modelId='amazon.titan-text-lite-v1', contentType='application/json', accept='application/json', body=json.dumps(payload) ) response_body = json.loads(response.get('body').read()) # Extract just the text from the Titan response structure print(response_body['results'][0]['outputText']) invoke_titan("What are the benefits of AWS Bedrock?")

---

Cost Estimate

This lab is extremely cost-effective and designed to be accessible:

• Amazon Q Developer: The Free Tier allows up to 50 chat interactions per month. This lab uses 1-2 interactions.
• Amazon Bedrock: On-demand inference is charged per 1,000 input/output tokens. The Titan Text Lite model costs fractions of a cent ($0.0003 per 1K input tokens). Completing this lab will cost less than $0.01.

---

Concept Review

To solidify your understanding, here is a breakdown of the AWS generative AI tools mentioned in this lab and the study material:

Hands-On Lab: Building with Amazon Bedrock and Amazon Q

Welcome to this guided hands-on lab! In this session, you will explore AWS's primary generative AI platforms. You will invoke foundation models using Amazon Bedrock, experiment with inference parameters, and interact with Amazon Q Developer to accelerate your cloud workflows.

---

Prerequisites

Before starting this lab, ensure you have the following:

• AWS Account: Access to an AWS account with AdministratorAccess or sufficient IAM permissions to use Amazon Bedrock and Amazon Q.
• CLI Tools: The AWS CLI installed and configured (aws configure) with your access keys.
• Region Selection: Set your default region to us-east-1 (N. Virginia) or us-west-2 (Oregon), as these regions have the broadest Amazon Bedrock model availability.
• Prior Knowledge: Basic familiarity with navigating the AWS Management Console and executing terminal commands.

---

Learning Objectives

By completing this lab, you will be able to:

1. Enable and configure foundation model access in Amazon Bedrock.
2. Invoke an Amazon Bedrock foundation model using both the AWS CLI and the AWS Management Console.
3. Adjust inference parameters (like Temperature) to alter model creativity.
4. Utilize Amazon Q Developer to ask AWS-specific architectural questions and generate code.

---

Architecture Overview

The following diagram illustrates the two distinct workflows you will execute in this lab:

---

Step-by-Step Instructions

Step 1: Request Model Access in Amazon Bedrock

Before you can use any foundation model in Amazon Bedrock, you must explicitly request access to it. This is a one-time setup step per region.

📸 Screenshot: The "Model access" page in the Amazon Bedrock console, showing "Manage model access".

▶Console alternative (Required for granting access)

1. Log in to the AWS Management Console and search for Amazon Bedrock.
2. In the left navigation pane, scroll down to Bedrock configurations and click Model access.
3. Click the Manage model access button at the top right.
4. Check the box next to Amazon Titan Text G1 - Lite (or similar available Titan Text model).
5. Scroll to the bottom and click Save changes.
6. Wait for the Access status to change to Access granted.

[!TIP] Model access is granted almost instantly for Amazon's own models (like Titan), but third-party models (like Anthropic Claude) may require submitting a use-case details form.

Step 2: Invoke a Model Using Amazon Bedrock

Now that you have access, we will send a prompt to the Amazon Titan model using the AWS CLI.

📸 Screenshot: Terminal window showing a successful JSON response from the Bedrock API.

▶Console alternative

1. In the Amazon Bedrock console, under Playgrounds in the left menu, select Text.
2. Click Select model, choose Amazon, and select Titan Text G1 - Lite.
3. Type Explain cloud computing in two sentences. in the prompt box.
4. Click Run to generate the response.

Step 3: Experiment with Temperature Settings

In Amazon Bedrock, adjusting the temperature affects the randomness of the model's responses. A temperature of 0.0 makes responses more deterministic, while higher values (up to 1.0) increase creativity and randomness.

Here is a visual representation of how temperature impacts token selection:

Step 4: Interact with Amazon Q Developer

Amazon Q Developer is tightly integrated into the AWS Console to help you understand services, troubleshoot errors, and write code.

📸 Screenshot: The Amazon Q chat panel docked on the right side of the AWS Management Console.

1. Open the AWS Management Console in your browser.
2. On the right side of the screen, click the Amazon Q icon (a colorful letter Q).
3. In the chat interface, type the following prompt:

[!NOTE] Amazon Q can inspect your active AWS environment (if permissions allow) to answer specific questions about your running resources and costs.

---

Checkpoints

After completing the steps, verify your progress:

1. Verify Bedrock CLI Access: Run aws bedrock list-foundation-models | grep amazon.titan-text-lite-v1. Expected Output: The console should return details of the Titan Text Lite model.

2. Verify Bedrock Invocation: Run cat response.json. Expected Output: A JSON string containing a "results" array with the generated text about cloud computing.

3. Verify Amazon Q Interaction: Expected Output: The Amazon Q side panel should provide an explanation of EC2 and a list/summary of your running instances.

---

Clean-Up / Teardown

While Amazon Bedrock charges on a pay-per-request basis (meaning no resources are actively left running), it is best practice to clean up local files. If you signed up for the Amazon Q Developer Pro tier ($19/month), remember to downgrade if you do not wish to be billed.

[!WARNING] Remember to run the teardown commands to avoid clutter and potential ongoing charges if premium subscriptions were activated.

To cancel Amazon Q Developer Pro (if enabled):

1. Navigate to the Amazon Q Developer console.
2. Select Subscriptions and choose to downgrade to the free tier.

---

Troubleshooting

Hands-On Lab: Getting Started with Amazon Bedrock and Amazon Q Developer

Welcome to this guided hands-on lab! In this session, you will explore AWS's primary generative AI platforms. You will learn how to enable and interact with Foundation Models (FMs) using Amazon Bedrock, understand the impact of inference parameters like Temperature, and use Amazon Q Developer as your intelligent coding and AWS assistant.

Prerequisites

Before you begin, ensure you have the following:

• An active AWS Account with Administrator or PowerUser access.
• The AWS CLI installed and configured on your local machine (aws configure).
• IAM Permissions allowing AmazonBedrockFullAccess.
• Basic familiarity with JSON and terminal commands.

[!WARNING] Some Amazon Bedrock Foundation Models incur charges based on the number of input and output tokens processed. Remember to follow the teardown instructions to remove any files, though simply having model access enabled does not accrue hourly charges.

Learning Objectives

By completing this lab, you will be able to:

1. Request and configure access to Foundation Models in Amazon Bedrock.
2. Invoke a generative AI model directly via the AWS CLI to generate text.
3. Adjust inference parameters (Temperature and Top P) to control model output.
4. Leverage Amazon Q Developer to ask AWS-specific architectural questions.

Architecture Overview

The following diagram illustrates how you will interact with both Amazon Bedrock and Amazon Q during this lab.

Step-by-Step Instructions

Step 1: Request Model Access in Amazon Bedrock

Before you can use a Foundation Model in Amazon Bedrock, you must explicitly request access to it. This ensures you review and accept the End User License Agreement (EULA) for the specific model provider.

▶Console alternative (REQUIRED for first-time setup)

1. Log in to the AWS Management Console and navigate to Amazon Bedrock.
2. In the left navigation pane, select Model access.
3. Click the Manage model access button.
4. Check the box next to Titan Text G1 - Lite (under the Amazon provider).
5. Click Request model access at the bottom of the page.
6. Wait for the Access status to change to Access granted.

📸 Screenshot: Model Access page showing "Access granted" next to Amazon Titan.

[!TIP] Amazon Titan models are typically granted instantly. Third-party models like Anthropic Claude may require additional use-case details to be submitted.

Step 2: Invoke a Foundation Model via CLI

Now that you have access, let's invoke the model to generate a response. We will pass a simple prompt asking the model to explain cloud computing.

▶Console alternative

1. In the Amazon Bedrock console, go to Playgrounds > Text.
2. Click Select model and choose Amazon > Titan Text G1 - Lite.
3. Type your prompt in the chat box.
4. Click Run to see the generated response.

Step 3: Experiment with Inference Parameters

Generative AI models use parameters like temperature and topP to control the randomness and creativity of the output.

Run the model again, but this time set the temperature to 0.0 for a highly deterministic response.

Step 4: Consult Amazon Q Developer

Amazon Q Developer is your AI assistant for software development and AWS knowledge. Let's use it to understand the invoke-model command we just ran.

▶Console alternative

1. Look for the Amazon Q icon on the right-hand sidebar of the AWS Management Console.
2. Open the chat panel.
3. Ask: "Why do I need to use --cli-binary-format raw-in-base64-out when calling Amazon Bedrock from the CLI?"
4. Review Amazon Q's response, which will explain that it prevents the AWS CLI from interpreting the binary output incorrectly, treating the JSON payload correctly.

📸 Screenshot: Amazon Q chat panel with the response and source citations.

Checkpoints

Verify your progress by running the following checks:

Checkpoint 1: Read the Model Output

Expected Result: A JSON response containing a results array with the generated text explaining Generative AI.

Checkpoint 2: Verify Deterministic Output

Expected Result: A JSON response containing a short 3-line poem (haiku) about cloud computing.

Clean-Up / Teardown

Because Amazon Bedrock models are serverless and charged per-token, you are not charged for idle time. However, it is good practice to clean up your local files.

[!WARNING] If you configured Provisioned Throughput for Amazon Bedrock (not covered in this basic lab), you must delete it in the console to avoid significant ongoing hourly charges.

Troubleshooting

Concept Review

Stretch Challenge

Want to test your skills? Try using Amazon Q Developer to write a Python script (using boto3) that automates Step 2. Then, run the Python script to invoke the amazon.titan-text-lite-v1 model without using the AWS CLI directly.

▶Show solution

pythonCopy

import boto3 import json client = boto3.client('bedrock-runtime', region_name='us-east-1') payload = { "inputText": "What are the benefits of AWS?", "textGenerationConfig": {"temperature": 0.7} } response = client.invoke_model( modelId='amazon.titan-text-lite-v1', contentType='application/json', accept='application/json', body=json.dumps(payload) ) response_body = json.loads(response['body'].read()) print(response_body['results'][0]['outputText'])

Curriculum Overview: Apply Natural Language Processing Services

[!IMPORTANT] Natural Language Processing (NLP) bridges the gap between human communication and machine understanding. In the AWS ecosystem, pre-trained AI services allow you to integrate powerful NLP capabilities without requiring deep data science expertise.

Prerequisites

Before diving into this curriculum, learners should have a foundational understanding of the following concepts:

Module Breakdown

This curriculum is structured to take you from foundational text processing theories to deploying fully managed AWS NLP services.

▶Click to expand: Why focus heavily on Preprocessing?

To deal with the complexities of human language, NLP involves considerable text processing. Before an AI model can effectively analyze data, the dataset must be cleaned. Techniques like removing punctuation and stopwords significantly reduce the computational load and improve accuracy by filtering out "noise."

Learning Objectives per Module

Module 1: NLP Foundations & Text Preprocessing

• Define Natural Language Processing and its role in modern AI applications.
• Differentiate between Stemming and Lemmatization in reducing words to their root forms.
• Apply lowercasing, stopword removal, and punctuation removal to unstructured text datasets.

[!TIP] Stemming vs. Lemmatization

• Stemming chops off the ends of words (e.g., "running" $\rightarrow$ "run"). It is fast but can be inaccurate.
• Lemmatization transforms a word to its dictionary root or lemma considering context. It is slower but more accurate.

Module 2: Evolution of NLP Models

• Trace the history of NLP from statistical methods to modern neural architectures.
• Understand how text is converted to numerical formats (Word2Vec, GloVe).
• Explain the role of Transformer architectures and self-attention mechanisms in Large Language Models (LLMs) like GPT and BERT.

$\vec{v}_{\text{king}} - \vec{v}_{\text{man}} + \vec{v}_{\text{woman}} \approx \vec{v}_{\text{queen}}$ Equation: A conceptual representation of semantic word embeddings.

Module 3: AWS Managed NLP Services

• Select the appropriate AWS service for specific text analysis tasks.
• Configure Amazon Comprehend to extract entities, key phrases, and sentiment.
• Design conversational interfaces using Amazon Lex.
• Implement intelligent document search using Amazon Kendra.

Module 4: Real-World Architecture & Integration

• Integrate multiple AWS AI services (e.g., Transcribe + Comprehend) into a cohesive pipeline.
• Assess business value and determine when a managed AI service is preferable to training a custom model.

Visual Anchors

AWS NLP Service Selection Flowchart

The Text Processing Pipeline

Success Metrics

To know you have mastered this curriculum, you should be able to achieve the following success metrics:

1. Architecture Design: Successfully draw an architecture diagram matching a business scenario to the correct AWS NLP services without referencing documentation.
2. Vocabulary Mastery: Clearly articulate the difference between intelligent document processing (IDP) and natural language processing (NLP).
   • Check: IDP automates data extraction from business documents (Amazon Textract), while NLP handles broader text-processing and linguistic comprehension tasks (Amazon Comprehend).
3. Exam Readiness: Consistently score 85%+ on practice questions related to the AWS Certified AI Practitioner (AIF-C01) NLP domain.
4. Hands-on Validation: Deploy a basic Amazon Lex chatbot that successfully triggers an AWS Lambda function to return a specific intent response.

Real-World Application

Natural Language Processing is no longer confined to academic research; it is actively transforming industries. Modern applications, such as customer service chatbots, now provide real-time responses by deploying sophisticated Transformer models in online inference settings.

Career Impact: Software engineers and application developers—even those without deep ML expertise—can leverage AWS's intuitive APIs to bring powerful NLP capabilities to market. This reduces development time from months to days.

Common Industry Use Cases:

• Healthcare: Using Amazon Comprehend Medical to extract medical ontologies and patient data from unstructured clinical notes.
• Customer Service: Modernizing contact centers by chaining Amazon Transcribe (speech-to-text) with Amazon Comprehend (sentiment analysis) to evaluate caller frustration in real-time.
• Global Commerce: Utilizing Amazon Translate and Amazon Polly to instantly localize product listings and provide multilingual accessibility features.

Prerequisites

Before diving into the application of Natural Language Processing (NLP) services on AWS, learners must possess foundational knowledge in the following areas:

• Cloud Fundamentals: Familiarity with basic AWS infrastructure, including IAM (Identity and Access Management) roles, permissions, and the AWS shared responsibility model.
• Basic AI/ML Concepts: Understanding of the differences between artificial intelligence, machine learning, deep learning, and generative AI.
• Data Literacy: Ability to differentiate between structured, unstructured, labeled, and unlabeled data. (NLP primarily deals with unstructured text data).
• Foundational Generative AI: Basic comprehension of generative models, tokens, embeddings, and transformer architectures.

Module Breakdown

This curriculum is structured to take you from the raw fundamentals of text processing to deploying enterprise-grade, AI-powered NLP services using AWS.

The Learning Path

Learning Objectives per Module

Module 1: The NLP Preprocessing Pipeline

• Objective: Prepare unstructured text data for machine learning models using standard linguistic techniques.
• Key Concept - Lemmatization: Transforming a word to its meaningful root (lemma) by removing affixes.
  • Real-World Example: A search engine converting the query "running shoes" to "run shoe" to match a broader set of relevant retail listings.
• Key Concept - Stemming: Chopping off the ends of words without considering context.
  • Real-World Example: An automated spam filter reducing "runner", "runs", and "running" all to the crude root "run" to quickly flag suspicious patterns.
• Key Concept - Stopword Removal: Filtering out words that add little semantic meaning (e.g., "the", "is", "at").
  • Real-World Example: Truncating a customer review from "The food is great" to "food great" to speed up database processing without losing the core sentiment.

Comparison: Stemming vs. Lemmatization

Module 2: Evolution of Text Representations

• Objective: Trace the historical progression of NLP and understand how modern AI interprets text.
• Techniques: Understand Bag-of-Words (BoW) and Term Frequency-Inverse Document Frequency (TF-IDF).
• Embeddings: Explain how models like Word2Vec and GloVe convert text into continuous mathematical vectors ($v \in \mathbb{R}^n$) to capture semantic relationships.
• Transformers: Describe how self-attention mechanisms in transformer architectures paved the way for Large Language Models (LLMs) like GPT and BERT.

Module 3: AWS Managed NLP Services

• Objective: Choose and implement the correct purpose-built AWS AI service for a specific business problem.
• Amazon Comprehend: Extract relationships, entities, and sentiment from unstructured text.
  • Real-World Example: Automatically tagging incoming support tickets as "Angry" or "Happy" to prioritize customer service responses.
• Amazon Lex: Build conversational interfaces (chatbots) using voice and text.
  • Real-World Example: Powering the self-service chatbot on a bank's website to help users check their account balances.
• Amazon Polly & Transcribe: Convert text-to-speech (Polly) and speech-to-text (Transcribe).
• Amazon Translate: Implement highly accurate, neural machine translation across different languages.

Module 4: Enterprise Search & LLMs

• Objective: Deploy advanced retrieval and generative solutions.
• Amazon Kendra: Provide highly accurate, AI-powered enterprise search.
  • Real-World Example: An internal corporate portal where employees can type natural language questions (e.g., "What is the maternity leave policy?") and receive exact answers extracted from hundreds of HR PDFs.
• Amazon Bedrock: Access and fine-tune foundation models to customize NLP applications using your organization's private data via Retrieval-Augmented Generation (RAG).

Success Metrics

To know you have mastered this curriculum, you should be able to:

1. Map Use Cases to AWS Services: Given a business scenario, correctly select between Comprehend, Lex, Textract, and Kendra with 100% accuracy.
2. Architect an NLP Pipeline: Successfully sketch a data flow from raw text collection $\rightarrow cleaning (lowercasing, stopword removal) \rightarrow$ inference via an AWS API.
3. Differentiate AI Categories: Clearly explain the difference between Natural Language Processing (understanding text) and Intelligent Document Processing (automating data extraction from visual documents like PDFs).
4. Evaluate Costs and Constraints: Assess the tradeoffs of using a simple pre-trained service (like Amazon Translate) versus fine-tuning a Large Language Model on Amazon Bedrock.

Real-World Application

Natural Language Processing is no longer an academic exercise; it is the backbone of modern enterprise automation and customer engagement.

Consider a globally distributed e-commerce company. They receive thousands of customer service calls and emails daily. By applying AWS NLP services, they can completely automate their pipeline:

Career Impact

For a software engineer or data professional, mastering these NLP services means you can integrate state-of-the-art AI into applications without needing a PhD in machine learning. Whether you are modernizing a contact center, generating dynamic product recommendations, or building automated compliance checks, AWS NLP tools drastically reduce the time-to-market for cutting-edge features.

AWS Certified AI Practitioner (AIF-C01) Curriculum Overview

This document provides a comprehensive roadmap for the AWS Certified AI Practitioner (AIF-C01) certification. This foundational-level credential validates your ability to recognize opportunities for AI/ML and implement them responsibly using AWS services.

## Prerequisites

Unlike Associate or Professional level certifications, the AIF-C01 is a Foundational exam.

• Prior Experience: No technical background or prior AWS experience is required. It is designed for individuals from both technical and non-technical backgrounds.
• Age Requirements: Candidates must be at least 13 years old (with parental consent for those 13–17).
• Recommended Knowledge: A basic understanding of IT services and their cloud-based applications is helpful but not mandatory.

AWS Certification Path

## Module Breakdown

The curriculum is structured into five core domains, moving from basic definitions to complex ethical and security frameworks.

## Learning Objectives per Module

Domain 1: Fundamentals of AI and Machine Learning

• Terminology: Define AI, ML, Deep Learning, and Neural Networks.
• Learning Types: Differentiate between Supervised (labeled data), Unsupervised (unlabeled data), and Reinforcement Learning (reward-based).
• The ML Lifecycle: Understand the steps from data collection and EDA (Exploratory Data Analysis) to model monitoring and retraining.

Domain 2: Fundamentals of Generative AI

• Core Concepts: Understand tokens, embeddings, and vectors.
• Architecture: Identify the role of Transformers as the backbone of modern Generative AI.
• Use Cases: Recognize applications in content creation, summarization, and code generation.

Domain 3: Applications of Foundation Models (FMs)

• RAG (Retrieval Augmented Generation): Explain how to ground models in private data using vector databases like Amazon OpenSearch.
• Prompt Engineering: Design effective prompts using context and logical reasoning steps.
• Service Selection: Choose between Amazon Bedrock (serverless FMs) and Amazon SageMaker (custom ML builds).

Domain 4: Guidelines for Responsible AI

• Bias & Fairness: Detect and mitigate bias using tools like SageMaker Clarify.
• Transparency: Use SageMaker Model Cards for documenting model intent and performance.
• Governance: Identify legal risks such as intellectual property infringement and hallucinations.

Domain 5: Security, Compliance, and Governance

• Data Protection: Implement encryption at rest and in transit.
• Infrastructure: Use AWS PrivateLink and IAM policies to restrict access to AI workloads.
• Governance Tools: Leverage AWS Audit Manager and CloudTrail for compliance tracking.

## Success Metrics

To earn the certification, candidates must demonstrate proficiency through a proctored exam.

[!IMPORTANT] Passing Score: 700 / 1,000 (Scaled Score)

Exam Format

• Questions: 65 total (50 scored, 15 unscored/experimental).
• Question Types: Multiple choice, multiple response, matching, and ordering.
• Duration: Typically 90–120 minutes.

## Real-World Application

This certification translates theoretical knowledge into practical business value.

• Business Leaders: Gain the vocabulary to lead AI initiatives and evaluate cost-benefit ratios (ROI).
• Developers: Learn to integrate pre-trained models via Amazon Bedrock without needing a PhD in Data Science.
• IT Professionals: Understand how to secure AI workloads and meet regulatory requirements (e.g., GDPR, HIPAA) within the cloud.

Example Use Case: Automated Customer Support

1. Service: Use Amazon Lex for the chatbot interface.
2. Intelligence: Use Amazon Bedrock to summarize previous customer interactions.
3. Governance: Use Bedrock Guardrails to ensure the bot doesn't provide harmful or biased advice.
4. Security: Use IAM to ensure the bot only accesses specific customer data.

[!TIP] Focus heavily on the "Responsibility" and "Security" domains (4 and 5), as these represent the "AWS way" of implementing AI, which is a major focus of the AIF-C01 exam.