Objectives
After completing this course, the learner will be able to:
■ Compare and contrast types of AI
■ Sketch the AI application deployment choices
■ Sketch the high-level LLM architecture
■ Provide an intuitive explanation of how LLMs work
■ List popular LLMs and communities
■ Create engineer prompts for various tasks
■ Sketch the process to enhance LLM with RAG and Vector databases
Outline
1. Intro to AI
1.1 Types of AI - Predictive, Generative, etc.
1.2 AI use cases in 5G - RAN and Core Network
1.3 Application deployment choices
2. Non-Generative AI
2.1 ML and DL defined
2.2 Types of learning
2.3 Predictive AI techniques
2.4 Types of ML/DL models
2.5 Use cases and associated techniques
Exercise: Match techniques with a 5G network problem
3. Introduction to GenAI and LLM
3.1 GenAI Application Architecture
3.2 How it works
3.3 Popular LLMs
3.4 Various LLM Communities
Exercise: Explain LLMs and how they work
4. Interacting with LLM
4.1 Ways to customize LLM responses
4.2 Incorporating your own data
4.3 APIs for LLMs
Exercise: Access LLM with APIs
5. Prompt Engineering with Playground
5.1 The magic of prompts
Exercise: Basic prompts with public LLM
5.2 Prompt anatomy
5.3 Zero-shot prompts
Exercise: Basic prompts
5.4 Few-shot learning
Exercise: Multi-shot prompts
5.5 Chain of Thought (CoT)
Exercise: CoT Exercise
5.6 Submission Cycle
Exercise: Submission Cycle Exercise
6. Engineering prompts for specific tasks
Exercise: Custom Task 1
Exercise: Custom Task 2
7. Retrieval Augmented Generation (RAG)
7.1 Why do need RAG?
7.2 RAG Architecture
7.3 VectorDB and tokenization
7.4 Importance of search criteria
Exercise: Using RAG with LLM
8. Fine Tuning
8.1 What is fine tuning?
8.2 How and when to fine tune?
9. Putting It All Together
10. Ask us about using your platforms/data