Open-Weight LLMs: Selection, Deployment, and Customization

Course details

Open-weight models are transforming how organizations build and deploy AI—offering more control, flexibility, and cost efficiency than closed APIs. In this course, instructor Laurence Moroney shows you how to navigate the rapidly evolving open model ecosystem and confidently apply these models in real-world scenarios. Explore different model types, where to find and run them, and how to choose the right setup based on your needs. Gain a practical understanding of customization approaches, including prompting, retrieval-augmented generation (RAG), and fine-tuning. By the end of this course, you’ll be equipped with a clear framework for selecting, deploying, and adapting open-weight models for your own projects. This course is an ideal fit for developers, data professionals, and technical practitioners who want more control over AI systems, as well as AI-curious builders looking to move beyond API-only workflows.

Learning objectives
Define what open-weight models are and distinguish them from open source and closed AI models.
Identify different types of AI models (text, vision, audio, code, and embeddings) and match them to real-world use cases.
Evaluate where and how to run open-weight models, including local environments, cloud providers, and managed platforms.
Apply a structured approach to customizing models using prompting, retrieval-augmented generation (RAG), and fine-tuning techniques.
Design practical AI solutions using open-weight models, based on constraints like cost, performance, and data sensitivity.

Concepts

Introduction

The power of local, private AI

Understanding the Open-Weight Landscape

What open weights actually means
The spectrum of model openness
Major model families
Licensing - What you can actually do

Finding and Running Open Models

Large language models
Vision-language models
Image generation models
Audio, speech, and code models
Embedding models
Model size vs. capability

Fine-Tuning and Customization

Why running locally matters
Hugging Face - The central hub
Running models locally with Ollama
llama.cpp and vLLM
Hardware reality check
Running models in the cloud
Choosing where to run

Applying Open Models in Real Scenarios

Fine-tuning overview
Why fine-tune at all
The customization spectrum
Low-rank adaptation (LoRA)
QLoRA and why it matters
The fine-tuning workflow
Choose your base model and train
Evaluate and deploy
Tools for fine-tuning

Putting it All Together

A decision framework
Scenario - Customer support chatbot
Scenario - Enterprise document classifier
Scenario - On-device assistant

Conclusion

The ecosystem is moving fast
Where to go from here