Building With Amazon Bedrock: A Foundational Workshop for Developers and Learners

Introduction

Amazon Bedrock has become one of the most practical services for teams that want to build generative AI applications without managing model infrastructure. Instead of provisioning GPU environments, hosting foundation models, or stitching together multiple providers from scratch, developers can access high-performing models through managed AWS APIs and focus on application design.

That sounds simple in theory, but most learners still face the same challenge: they can call a model, yet they do not fully understand what is happening underneath. Terms such as tokens, embeddings, temperature, streaming, tool use, and multi-turn messaging appear quickly, and without a guided learning path, the concepts often remain fragmented.

This foundational workshop was designed to solve that problem.

It provides a structured, lab-based introduction to Amazon Bedrock so that students, engineers, and self-learners can move from basic model access to practical implementation patterns with confidence.

Repository reference:

• GitHub repository: https://github.com/TuWienProjects/aws-cloud-workshop3-bedrock

Direct documentation path inside the repository:

• Foundational track: docs/workshops/workshop-03-bedrock/foundational/README.md

Why a Foundational Bedrock Workshop Matters

Many developers start working with large language models by experimenting with prompts in a playground. While that is a useful first step, production-oriented understanding requires more than prompt experimentation.

You also need to understand:

• How Amazon Bedrock exposes models through different API styles

• When to use direct invocation versus conversational interfaces

• How inference parameters affect determinism, creativity, and cost

• Why streaming improves user experience in interactive systems

• How embeddings support semantic retrieval and search

• How lightweight UI frameworks such as Streamlit accelerate prototyping

This workshop establishes that foundation before learners move into advanced agentic workflows.

Who This Workshop Is For

This workshop is suitable for:

• Beginners who want a structured introduction to Amazon Bedrock

• Developers who already know Python and want Bedrock implementation examples

• Students preparing coursework, labs, or technical blog content

• Engineers building proof-of-concept generative AI applications on AWS

No prior machine learning specialization is required. A basic understanding of Python, command-line tooling, and AWS access is enough to get started.

What Learners Will Gain

By completing the foundational workshop, learners should be able to:

• Explain what Amazon Bedrock is and why it matters in modern AI application development

• Use Amazon Bedrock from both the AWS Console and Python code

• Understand the difference between InvokeModel and Converse

• Control response behavior with inference parameters such as temperature

• Implement streaming responses for responsive user experiences

• Generate and compare text embeddings for semantic similarity use cases

• Build a simple Streamlit-based prototype for demonstrating AI functionality

Workshop Setup and Learning Path

The workshop is designed for learners using either:

• A workshop-provided AWS environment

• Their own AWS account and local Python setup

Before starting, learners should confirm:

• Amazon Bedrock access is enabled in a supported AWS Region

• The required foundation models are available in the account

• Python and boto3 are installed if running locally

• AWS credentials are configured correctly

This matters because environment preparation is often the first source of friction in generative AI labs. A professional workshop should reduce setup ambiguity, not increase it.

Lab-by-Lab Breakdown of the Foundational Track

The foundational track is organized as a progression from conceptual orientation to practical implementation.

Lab F-1: Bedrock Introduction

This lab introduces Amazon Bedrock as a managed service for accessing foundation models through a unified AWS experience.

Learners begin by understanding:

• What a foundation model is

• Why a managed service model is operationally valuable

• How Bedrock simplifies experimentation and application development

• How to navigate the console and explore supported model families

This lab is especially important because it gives non-specialists a clean conceptual anchor before they begin writing code.

Lab F-2: InvokeModel API

This is the first programmatic step.

Learners use Python and boto3 to make a direct model invocation. This introduces:

• Bedrock runtime client usage

• JSON request payloads

• Basic response parsing

• The mindset of treating model access as application infrastructure rather than a novelty

For developers, this is where the workshop starts to feel practical.

Lab F-3: Converse API

Once learners understand direct invocation, the workshop introduces the Converse API, which is a better fit for structured conversational interactions.

This lab frames key ideas such as:

• Message-based interaction models

• Consistent request and response structure

• Multi-turn conversation design

• The role of system instructions in shaping output behavior

This is a critical transition point because modern generative AI applications are rarely single-shot systems.

Lab F-4: Inference Parameters

A good AI workshop should not stop at “send prompt, receive response.”

This lab introduces inference configuration and teaches learners how to reason about:

• Output length control

• Predictability versus variability

• Cost implications of token generation

• Model behavior tuning for different use cases

This is where learners begin moving from experimentation to intentional engineering.

Lab F-5: Temperature

Temperature deserves focused treatment because it is one of the most misunderstood model parameters.

This lab helps learners understand:

• Why low-temperature outputs are more deterministic

• Why higher temperature increases diversity and creativity

• Why the correct setting depends on business context

• Why creative generation and process-oriented generation should be treated differently

For anyone building real applications, this distinction is essential.

Lab F-6: Tool Use

This lab introduces one of the most important conceptual bridges in modern AI systems: allowing a model to request structured external actions.

Learners begin to understand:

• Why models should not be treated as isolated text generators

• How tool schemas describe callable functionality

• How applications can interpret model requests and execute deterministic logic

• Why tool use is foundational to automation and intelligent workflows

This lab also acts as a conceptual bridge to the agentic workshop.

Lab F-7: Streaming API

Streaming improves the perceived responsiveness of AI systems.

Instead of waiting for the entire output to finish, learners see how to:

• Receive incremental response chunks

• Render content progressively

• Improve user experience in conversational and interactive interfaces

This is especially relevant for chatbots, assistants, and live demos.

Lab F-8: Embeddings

Embeddings are one of the most important building blocks in retrieval-augmented generation and semantic search.

This lab introduces:

• Text-to-vector conversion

• Semantic similarity

• Cosine similarity as a comparison technique

• The connection between embeddings and search-style AI systems

For learners who want to progress into RAG systems later, this lab is foundational.

Lab F-9: Intro to Streamlit

A strong workshop does not only teach model calls. It also teaches how to present working ideas.

This lab shows learners how to use Streamlit to:

• Create quick interfaces with Python

• Capture user input

• Display model outputs visually

• Build demo-friendly prototypes without full front-end complexity

This is particularly useful for hackathons, coursework, early proofs of concept, and stakeholder communication.

Why This Workshop Is Valuable for Real Projects

The strongest quality of this foundational track is that it does not treat Amazon Bedrock as an abstract service. It treats it as a practical platform for building software.

That distinction matters.

A professional learning experience should help readers understand both the conceptual model and the engineering model:

• Conceptually, what the service is doing

• Programmatically, how to interact with it

• Architecturally, where it fits in a broader application

• Operationally, what trade-offs affect usability, performance, and maintainability

That is what makes this workshop useful beyond a single lab session.

Recommended Reader Journey

If you are publishing this workshop to a broad audience, the best reader path is:

1. Start with the foundational track overview

2. Complete the API-focused labs in order

3. Revisit embeddings and Streamlit for applied experimentation

4. Continue into the agentic track once the Bedrock basics feel natural

This sequencing keeps the learning curve progressive rather than overwhelming.

Final Thoughts

Amazon Bedrock lowers the barrier to building with foundation models, but meaningful understanding still requires careful practice. A well-designed foundational workshop fills that gap by converting abstract AI terminology into concrete developer capability.

If your goal is to learn Amazon Bedrock properly, teach it professionally, or create strong blog-linked workshop material, this foundational track provides the right starting point.

For the complete hands-on material, code structure, and workshop documentation, refer to the repository:

• https://github.com/TuWienProjects/aws-cloud-workshop3-bedrock

And when you are ready to move beyond model invocation into tool-driven reasoning and structured AI workflows, continue with the companion agentic workshop.