Build Local AI Agents with Gemma and Ollama

Course details

Master the cutting-edge world of AI development with Google's powerful open-source Gemma 3 model in this comprehensive hands-on course. Instructor Laurence Moroney shows you how to build an end-to-end concierge agent that plans, reasons, reflects, and uses tools, leveraging Ollama to keep your application running. Along the way, find out how to create a smart application that uses LLMs to determine your desired task, and execute using tools to search the web, browse and understand websites, send email, and more.

This course is designed for Python developers who want to add AI capabilities to their skill set, as well as tech-savvy entrepreneurs or product managers looking to get hands-on experience building, deploying, and managing intelligent applications.

Learning objectives
Understand the agentic lifecycle of planning, executing, and reflecting.
Implement tools for an agent to use for searching the web, browsing websites, and sending email.
Build an application on Gemma that provides a local, private, personal concierge.
Serve local Gemma models from your device.

Concepts

Introduction

Build an agent today with Gemma 3

Getting Started - Gemma 3 Setup

Part 1 setup - Gemma 3 environment
Part 2 setup - Ollama installation to run Gemma 3 locally
Overview - Your first Gemma 3 project

Designing and Planning Your Agent

Design your agent for this project
Insert an API into an agentic workflow
Reflect on the results and continue

Implementing the Tools for the Agent to Use

Overview - The tools used in this agentic workflow
Search the web for content
Browse specific sites
Report results back to the user

Creating Prompts for the Agentic Workflow

Overview - Create prompts for the agentic workflow
Design a prompt to turn input into actions
Design a prompt to create commands for the site browsing
Design a prompt to fact-check results from the websites
Design a prompt to ascertain quality of results

Finishing Your App

Almost done The flow that builds the application
Finish the code

Extending Your App to Be Multimodal

Work with multimodal input
Interpret the contents of an image
Update the agentic workflow to take image contents
Refine the workflow to refine results

Conclusion

What's been covered in the course

Build Local AI Agents with Gemma and Ollama

Authors

Laurence Moroney