Large language models are now the core of most AI applications, and getting good results from them is a skill of its own. This module covers what LLMs are and how they're built, from tokens and pretraining to fine-tuning into a chat assistant, then how to steer them with prompt engineering: zero-shot and few-shot prompts, breaking tasks into steps, role-based prompting, and handling hallucinations.
The prompting lessons are practical. They include using AI for coding, debugging, documentation and the job search, along with best practices and common pitfalls.
These courses are long, so treat them as a reference: watch the parts you need, then come back for the rest. When you're done, continue on to LLM APIs & Structured Output.
What prompt engineering is, a short introduction to AI and language models, the prompt engineering mindset, using GPT-4, best practices, zero-shot and few-shot prompts, AI hallucinations, and text embeddings.
Controlling GPT's responses and breaking tasks into smaller steps, role-based prompting and AI pitfalls, AI-assisted coding, debugging and documentation, and using language models for the job search.
How an LLM is built end to end: preparing data, building a Byte Pair Encoding tokenizer, the transformer architecture and attention, pretraining, then instruction tuning with LoRA to produce a chat assistant.