Introduction to Prompt Engineering Deep Dive

Let us start with something simple.

If you’ve ever used ChatGPT or Gemini casually, you’ve probably typed something like:

“Explain photosynthesis.”

Better prompts don’t come from longer instructions.
They come from clearer thinking.

Introduction to Structured outputs and Prompt Formatting

Let’s do something a little different.

Instead of starting with definitions, I want to show you a problem.

If there’s one thing to remember, it’s this:

AI responses become useful only when they become predictable.

And predictability comes from structured prompting.

Introduction to Prompt Chaining for Multi-Step AI Workflows

Let’s build something slightly more real today.

So far, every time we’ve used the model, we’ve done one thing:

We send a prompt → we get a response → done.

one output becomes the next input.

Then we structured the outputs.

And finally, we built a simple multi-step workflow.

This is how developers move from simple prompts to real AI systems.

Introduction to Understanding Tokens

In the previous lesson, we made our first Gemini API call using Python.

We sent a prompt to the model, and the model generated a response.

In the next lesson, we’ll go deeper into prompt engineering, where we’ll learn how to structure prompts so the model produces more useful and consistent outputs.

Introduction to first Gemini API call

we’re going to make our first Gemini API call using Python, and we’ll do it inside Google Colab, which is one of the easiest ways to write and run Python code in the cloud.

By the end of this lesson, you’ll be able to:

• open a Colab notebook
  • install the Gemini Python library
  • connect to the Gemini API
  • send your first prompt from Python
  • receive the model’s response in your code

Once you understand this step, you’ve officially moved from testing prompts to building AI-powered applications.

Step 1: Open Google Colab

Start by opening your browser and going to Google Colab.

https://colab.research.google.com

Google Colab is an online Python environment where you can write and run code without installing anything on your computer.

Click New Notebook.

You should now see a blank notebook with a code cell ready.

This notebook will act as our small development environment where we can interact with the Gemini API.

Step 2: Install the Gemini Python Library

Before our code can communicate with Gemini, we need to install the Python library that allows our notebook to interact with the API.

In the first code cell, type the following command:

!pip install -q google-generativeai

Then run the cell.

You can run a cell by clicking the play button on the left side of the cell.

This installs the official Gemini Python SDK.

If you’re curious about what this library does or want to explore more examples later, you can visit the official documentation here:

https://ai.google.dev/docs

Once the installation finishes, we’re ready to connect to the Gemini API.

Step 3: Generate a Gemini API Key

Before your code can access Gemini, Google needs to know who is making the request.

That’s done using something called an API key.

Open a new browser tab and go to:

https://aistudio.google.com/app/apikey

If this is your first time, click Create API Key.

Google will generate a key for you.

Copy that key and keep it private. Think of it like a password that allows your program to access the AI service.

Once you’ve copied the key, go back to your Colab notebook.

Step 4: Configure the API in Python

In a new code cell, paste the following code:

import google.generativeai as genai

genai.configure(api_key=”PASTE_YOUR_API_KEY_HERE”)

Replace the text inside the quotes with the API key you copied.

This code imports the Gemini library and configures it to use your API key.

Run the cell.

You won’t see any output yet, but your notebook is now connected to the Gemini service.

Step 5: Select a Gemini Model

Next, we need to choose which model we want to use.

Add another code cell and type:

model = genai.GenerativeModel(“gemini-1.5-flash”)

Gemini offers several models with different capabilities.

For most development tasks, Gemini Flash is a great starting point because it is fast and efficient.

If you want to explore the available models later, you can see them here:

https://ai.google.dev/gemini-api/docs/models

Run the cell.

Now our notebook knows which model it will send requests to.

Step 6: Send Your First Prompt

Now we’re ready to send our first prompt from Python.

In a new cell, type the following:

response = model.generate_content(

“Explain cloud computing in simple terms.”

)

Run the cell.

This line sends a request to the Gemini model with the instruction we provided.

But we haven’t displayed the response yet.

So in the next line, add:

print(response.text)

Run the cell again.

You should now see the model’s response printed directly inside your notebook.

That response is coming from the Gemini model in real time.

Step 7: Try Another Prompt

Let’s try another example so you can see how flexible this interaction is.

Replace the prompt with something more structured:

response = model.generate_content(

“Give me three real-world applications of artificial intelligence in healthcare.”

)

print(response.text)

Run the cell again.

You’ll see Gemini generate a list of examples.

At this point, you can experiment with any prompt you want.

Your code is now able to communicate directly with the model.

Understanding What Just Happened

Let’s briefly recap what we just did.

Our notebook followed four simple steps.

First, we installed the Gemini Python library.

Second, we connected our notebook to the Gemini API using an API key.

Third, we selected a model.

And fourth, we sent a prompt and printed the response.

This basic interaction is the core of almost every generative AI application.

Whether you’re building a chatbot, a writing assistant, a research tool, or a document analyzer, the application is essentially sending prompts to a model and receiving responses.

How This Fits into Real AI Applications

In real systems, this code usually sits inside a backend service.

For example, a user might type a question into a web app.

That app sends the question to a server.

The server then sends that input to the Gemini API.

The model generates a response.

And the application displays that response back to the user.

So even though this example looks simple, it represents the core interaction that powers modern AI applications.

Summary – Key Takeaways

You learned how to:

• open a Google Colab notebook
• install the Gemini Python library
• generate an API key
• connect Python to the Gemini API
• send prompts from Python
• display the model’s response

This is the point where prompt experimentation turns into real development.

In the next lesson, we’ll start turning this basic API call into something more useful by building a small AI-powered tool.

Introduction to Vertex AI Studio Playground

Vertex AI Studio is where developers experiment with Gemini models before building real AI applications.

In this video, you will learn how to:

• open Vertex AI Studio
• understand the Playground interface
• test Gemini chat prompts
• generate and explain code
• explore multimodal capabilities like image input

By the end of this lesson, you will know how developers test ideas in Vertex AI before integrating AI into applications.

Open Google Cloud

First, open your browser and go to the Google Cloud Console.

https://console.cloud.google.com/

Log in using your Google account.

If this is your first time using Google Cloud, you may need to create a project.

Google provides a simple guide for beginners here:

https://cloud.google.com/docs/get-started

Once you are inside the console, look at the search bar at the top.

Type Vertex AI.

Click Vertex AI from the results.

Now look at the menu on the left side.

Click Vertex AI Studio.

If you want to open it directly in the future, you can bookmark this page:

https://console.cloud.google.com/vertex-ai/generative

You are now inside the workspace where you can test Gemini models.

What Vertex AI Studio Is

Vertex AI Studio is a testing environment for generative AI models.

Developers use it to:

• experiment with prompts
• explore how the model responds
• adjust model settings
• validate ideas before writing application code

Instead of building an entire application first, developers test their ideas here.

Once they find prompts and workflows that work well, they integrate them into real applications using APIs.

You can learn more about Vertex AI generative tools here:

https://cloud.google.com/vertex-ai/docs/generative-ai/learn/overview

Understanding the Playground Interface

Let’s look at the main parts of the Playground interface.

You will usually see four key areas.

First is the model selector.
This allows you to choose the Gemini model you want to use.

Second is the prompt input area.
This is where you type your instructions.

Third is the model settings panel.
Here you can adjust parameters like creativity and output behavior.

Fourth is the output section.
This is where the model’s response appears after you run a prompt.

The workflow is simple.

Choose a model, write a prompt, run it, and review the response.

This is how developers test ideas before building applications.

Testing Gemini Chat Mode

Let’s start by testing chat mode.

Chat mode is useful for tasks like:

• answering questions
• summarizing information
• drafting text
• brainstorming ideas

In the prompt area, type the following instruction.

Explain cloud computing in simple terms for a college student.

Click Run.

The model generates a response.

Now let’s improve the prompt.

Developers rarely stop at the first prompt. They improve it step by step.

Try this prompt instead.

Explain cloud computing in simple terms for a college student. Use short paragraphs and one real world example.

Run the prompt again.

Notice how the response becomes clearer and more structured.

This process is called prompt iteration.

Developers refine prompts in the Playground until they get reliable output.

You can learn more about prompt engineering here:

https://cloud.google.com/vertex-ai/docs/generative-ai/prompt-engineering

Testing Code Generation

Gemini can also assist with programming tasks.

It can help with:

• generating code
• explaining code
• debugging code
• converting code between languages

Let’s try a simple example.

Type the following prompt.

Write a Python function that returns the average of a list of numbers.

Run the prompt.

Gemini generates the Python code.

Now let’s try a different type of coding request.

Explain this Python function line by line:

def average(nums):
return sum(nums) / len(nums)

Run the prompt again.

Now Gemini explains the code step by step.

This is useful for developers building coding assistants or for learners who want help understanding code.

You can explore the Gemini API documentation here:

https://ai.google.dev/docs

Testing Multimodal Input

Modern AI models can understand more than text.

They can also work with images and other types of input.

This capability is called multimodal AI.

Depending on the interface and model, the Playground may allow you to upload an image.

Upload a simple image or screenshot.

Now enter this prompt.

Describe what is happening in this image in simple language.

Run the prompt.

The model now combines the image and the text instruction to produce a response.

Multimodal AI is useful for many real-world applications, including:

• analyzing charts and diagrams
• describing images
• reading screenshots
• document analysis

You can learn more about Gemini’s multimodal capabilities here:

https://ai.google.dev/gemini-api/docs/multimodal

Understanding Model Settings

You may notice some model settings on the side panel.

One important setting is temperature.

Temperature controls how creative or varied the response is.

Lower temperature usually produces:

• more focused answers
• more consistent responses

Higher temperature produces:

• more creative responses
• more variation in wording

For enterprise tools like customer support systems or document analysis, developers often use lower temperature.

For creative writing or brainstorming tools, higher temperature may be useful.

More details about model parameters can be found here:

https://cloud.google.com/vertex-ai/docs/generative-ai/model-reference

How Developers Use the Playground

Here is the typical workflow developers follow when using Vertex AI Studio.

First, they define a use case.

Second, they write a prompt.

Third, they run the prompt and review the response.

Fourth, they refine the prompt to improve the output.

They repeat this process until the response is reliable and useful.

Once the prompt works well, they connect the model to their application using APIs.

This approach saves time and helps developers build better AI applications.

Summary – Key Takeaways

You learned how to:

• open Vertex AI Studio in Google Cloud
• understand the Playground interface
• test Gemini chat prompts
• generate and explain code
• experiment with multimodal inputs
• adjust model settings

Vertex AI Studio is the place where developers experiment and test ideas before building AI-powered applications.

In the next lesson, we will move from testing prompts to making our first Gemini API call using Python and Google Colab.

Introduction to Gemini – Google’s Generative AI Models

Gemini is Google’s latest family of large language models (LLMs), developed to compete with and improve upon existing models like OpenAI’s GPT series. Gemini powers many of Google’s generative AI experiences, from search enhancements to workspace assistants, and it is also available to developers.

Gemini is capable of:

• Understanding and generating text
• Summarizing documents
• Writing code
• Translating content
• Answering questions based on context
• Handling multimodal input (text, images, audio)

You can think of Gemini as Google’s most advanced “thinking engine” — designed not only to answer questions but to help developers build AI-powered applications.

What is PaLM and How is it Related to Gemini?

Before Gemini, Google developed a family of models called PaLM (Pathways Language Model). These models, like PaLM 2, were used in early versions of Bard and in tools inside Google Workspace (like Gmail and Docs).

In late 2023, Google transitioned from PaLM to Gemini:

• PaLM → Gemini 1.0
• Gemini 1.5 models (released in 2024) offer better reasoning and longer context windows
• These models are now powering tools inside Android, Chrome, and Gemini apps

So if you used Bard or Workspace AI features before, you likely interacted with PaLM. Now, Gemini has replaced it as the flagship model line.

What is Vertex AI Studio?

Vertex AI Studio is Google Cloud’s interface for experimenting with generative AI models like Gemini. It is a developer-friendly playground where you can test prompts, fine-tune settings, and generate content using a web interface — no installation or coding required.

Key features of Vertex AI Studio:

• Prompt writing and testing in a browser
• Multiple model options (Gemini 1.0 Pro, Gemini 1.5 Pro, CodeGemma, etc.)
• Exportable code snippets for Python or REST API
• Project-saving for versioning
• Integration with Vertex AI Workbench (for building apps)

How Do These Tools Work Together?

Here is how everything connects:

Tool / Concept	Role or Purpose
Gemini	The actual AI model that performs the task
PaLM	The previous generation model family
Vertex AI Studio	The playground to test Gemini prompts
Vertex AI API	Lets you integrate Gemini into your own apps using code
Google Cloud Console	The central hub to manage access and billing

When you write a prompt in Vertex AI Studio, the system sends it to a Gemini model and returns the response in real time. If you build a web app or script, your code talks to the same Gemini model through the Vertex AI API.

Where Can You Access Gemini?

There are multiple access points for Gemini:

1. AI Studio — Direct interface for prompt testing
2. Makersuite — Lightweight version of AI Studio for experimentation
3. Google Workspace — In tools like Gmail, Docs, and Sheets via Duet AI or Gemini
4. Google Cloud SDK & API — For app development and integration
5. Gemini app — On Android or iOS (consumer-friendly version)

As a developer, your focus will primarily be on AI Studio and Vertex AI API, which allow you to test and integrate your apps with Google’s LLMs.

Summary – Key Takeaways

• Gemini is Google’s most advanced LLM, replacing the earlier PaLM family.
• Vertex AI Studio is the main tool to explore and test Gemini’s capabilities using prompts.
• Developers can use Gemini via the cloud interface, SDK, or API.
• Understanding the relationship between Gemini, PaLM, and Vertex AI Studio helps you build smarter applications using Google Cloud.

What is Google Cloud Platform?

Google Cloud Platform (GCP) is a suite of cloud computing services provided by Google. It allows you to run code, store data, deploy applications, and access powerful machine learning tools — including the Vertex AI platform, which powers tools like Gemini.

Why Use Free Coding Platforms?

As an AI developer, you need tools that allow you to write, run, and share code easily. Especially when you are just starting, it is best to use platforms that are free and do not require you to set up complex infrastructure.

Three powerful tools used by AI developers worldwide are:

• Google Colab – for notebooks and experiments
• Streamlit – for building and deploying web apps
• Replit – for real-time, browser-based coding and collaboration

Each tool serves a unique role in your development journey.

Google Colab – Run Python in the Cloud

Google Colab is an online notebook environment built by Google. It allows you to write and run Python code in the browser. It is especially popular in data science and AI communities.

Key features:

• Free cloud-based access to CPU and GPU
• No installation required
• Notebook format allows for combining code, text, and images
• Easy to share and collaborate

Use cases:

• Test AI models and APIs
• Write and refine prompts
• Try sample code before deploying

Getting started:

1. Go to https://colab.research.google.com
2. Click New Notebook
3. Write a line of Python, like print(“Hello Generative AI”)
4. Press Shift + Enter to run

Streamlit – Build and Share AI-Powered Web Apps

Streamlit is a Python framework that turns your Python scripts into interactive web applications. You can build apps for text generation, form submissions, file uploads, and much more — all using simple Python code.

Key features:

• Build dashboards and tools without HTML or JavaScript
• Deploy directly from GitHub using https://share.streamlit.io
• Supports input forms, buttons, sliders, and markdown

Use cases:

• Build an AI assistant
• Create a resume analyzer or Q&A bot
• Host your app with a public link

Getting started:

1. Sign up at https://streamlit.io
2. Install Streamlit (if using locally): pip install streamlit
3. Write a script:

import streamlit as st

st.title(“Welcome to my AI app”)

1. Run locally using: streamlit run your_script.py
   Or deploy from GitHub for public sharing

Replit – Code Together, in the Cloud

Replit is an in-browser code editor that supports multiple programming languages. It allows real-time collaboration, similar to Google Docs but for code.

Key features:

• Instant coding without setup
• Run and preview Python, HTML, JavaScript, and more
• Free hosting for small projects
• GitHub integration for version control

Use cases:

• Collaborate with classmates or teammates
• Prototype small AI tools
• Host chatbot APIs and widgets

Getting started:

1. Visit https://replit.com
2. Sign up and click Create Repl
3. Choose Python or HTML + JS
4. Start coding and click Run

When to Use Which Platform?

Platform	Best For	Requires Installation?	GitHub Integration?
Google Colab	Testing code, writing notebooks	No	Yes
Streamlit	Building shareable apps	Yes (for local use)	Yes
Replit	Quick prototyping and hosting	No	Yes

Together, these tools allow you to build a complete project — from idea to code to deployment — without ever leaving your browser.

Summary – Key Takeaways

• Google Colab is perfect for experimenting with code and writing in a notebook style.
• Streamlit is best for building and deploying interactive web apps with AI.
• Replit is great for real-time collaboration and hosting small projects.
• All three platforms work well with GitHub and are ideal for Generative AI learners.

What is a Prompt?

In Generative AI, a prompt is the instruction or input you give to a model. It could be a question, a command, or even a few lines of example content. The model responds by generating new text, code, summaries, or other content based on what you ask.

Prompts are how we “communicate” with large language models. Just like giving good directions to a friend, the better your prompt, the better the response.

Why Does Prompt Engineering Matter?

A small change in a prompt can lead to very different results. Prompt engineering is the practice of designing and testing different types of input to get the most useful or accurate responses from the model.

Here are three examples of the same task, writing a product description:

Prompt A: Write a description for a water bottle.

Prompt B: Write a short, energetic product description for a reusable water bottle targeted at fitness enthusiasts.

Prompt C: Write a 3-sentence product description for a stainless steel reusable water bottle. Highlight its eco-friendliness, durability, and portability.

All three prompts ask for the same thing, but the more specific prompts will likely produce better results. Prompt engineering is about learning how to make your intent clear and structured.

Components of a Good Prompt

When writing prompts, consider the following parts:

1. Instruction: Clearly state what you want the model to do. Example: “Summarize this article.”
2. Context: Give background or details. Example: “This article is about climate change solutions.”
3. Format: Tell the model how to structure the response. Example: “Write in bullet points.”
4. Tone or Style (optional): Add stylistic cues. Example: “Make it sound friendly and informal.”

You do not always need all four parts, but combining them can make your prompt more effective.

Trying Prompts in ChatGPT or Gemini

You can test your prompts using free tools like:

• ChatGPT (https://chat.openai.com)
• Google Gemini (https://aistudio.google.com/app/prompts)

Try typing different instructions and observe how the output changes when you:

• Add or remove context
• Change the tone
• Switch from general to specific requests

For example, start with:

“Summarize the article below.”

Then try:

“Summarize the article below in 3 bullet points, keeping the tone friendly and the language simple for high school students.”

Notice how clarity improves the output.

Summary – Key Takeaways

• Prompts are how you give instructions to an AI model.
• Better prompts lead to better responses.
• Prompt engineering helps you get useful, relevant, and accurate answers.
• Tools like ChatGPT and Gemini allow you to experiment freely with prompt structures.