Building with Llama 4
Course Details
Course Curriculum & Description
Course Overview
Building with Llama 4 is a beginner-level short course developed by DeepLearning.AI in collaboration with Meta and taught by Amit Sangani, Senior Director of Partner Engineering at Meta.
The course introduces Meta's Llama 4 family of models and demonstrates how developers can use its Mixture-of-Experts (MoE) architecture, multimodal capabilities and long-context processing to build AI applications.
Through practical lessons, learners explore Llama 4 Maverick and Scout, work with Meta's official Llama API and Together.ai, and experiment with image grounding, multilingual applications, prompt optimization and synthetic data generation.
What You'll Learn
Understand the Llama 4 model family and its Mixture-of-Experts architecture. Learn to select and use Llama 4 models, build applications through an API, reason across multiple images, identify objects and their bounding boxes, and process large text collections. Explore techniques for improving system prompts and generating synthetic datasets for model fine-tuning.
Course Curriculum
1. Important Note — 10 minutes (Reading)
Review important course information before beginning the lessons.
2. Introduction — 3 minutes
Meet the instructor and understand the course objectives, structure and practical applications of Llama 4.
3. Overview of Llama 4 — 6 minutes
Explore the evolution of Llama models, the Llama 4 family and the Mixture-of-Experts architecture. Understand the characteristics of Maverick and Scout.
4. Quickstart with Llama 4 and API — 6 minutes
Learn how to access Llama 4 through Meta's API and build a multilingual translator chatbot.
5. Image Grounding — 9 minutes
Explore multimodal image reasoning, object identification and bounding-box detection. Experiment with generating executable code from user interface screenshots.
6. Llama 4 Prompt Format — 8 minutes
Understand Llama 4's special tokens and raw prompt structures for text-only and multimodal interactions.
7. Long-Context Understanding — 7 minutes
Learn to process and query large documents, research papers, books and code repositories using Llama 4's long-context capabilities.
8. Prompt Optimization Tool — 10 minutes
Use Llama's prompt optimization tool to improve system prompts for applications such as sentiment analysis and text categorization.
9. Synthetic Data Kit — 7 minutes
Learn to ingest, generate, curate and save synthetic datasets for training and fine-tuning.
10. Conclusion — 1 minute
Review the capabilities explored throughout the course and their potential applications.
Graded Quiz — 10 minutes
Complete the course assessment to evaluate your understanding of the material.
Models and Technologies Covered
Llama 4 Maverick, Llama 4 Scout, Mixture-of-Experts architecture, Meta Llama API, Together.ai, multimodal prompting, image grounding, long-context processing, prompt optimization and synthetic data generation.
Who Should Enroll?
This course is designed for developers, AI enthusiasts, students and anyone interested in gaining practical experience with the Llama 4 family of models. Its beginner-level structure introduces the models and demonstrates their capabilities through application-focused lessons.
Course Completion
The course includes video lessons, reading material and a graded quiz. Learners with DeepLearning.AI Pro can earn a course accomplishment, subject to the platform's applicable requirements.
What You Will Learn
- Understand Llama 4's Mixture-of-Experts architecture
- Build AI applications using the Llama API
- Develop multilingual AI chatbots
- Perform multimodal image reasoning
- Identify objects using image grounding
- Understand Llama 4 prompt formatting
- Process long-context documents and code repositories
- Optimize system prompts automatically
- Generate synthetic datasets for model fine-tuning
- Work with Llama 4 Maverick and Scout
Course Prerequisites
No formal prerequisites are specified in the supplied course description. Basic familiarity with programming, APIs and generative AI concepts may be helpful for the practical exercises.
Course Highlights
Beginner-friendly short course, Developed in collaboration with Meta, Taught by a Meta engineering leader, Hands-on Llama 4 API exercises, Multimodal image reasoning and grounding, Long-context processing, Prompt optimization, Synthetic data generation, Graded assessment, Pro accomplishment available