Beginner to builder · 34 lessons · ~8 hours

AI Engineering: Build Real AI Apps

The skills behind every AI product: how models work, testing them, prompts, RAG, agents, finetuning and shipping. Based on ideas from Chip Huyen's AI Engineering.

What you walk away with

Explain how AI models produce answers
Test an AI app with a real eval set
Build a RAG or agent plan for a real problem

Syllabus

1. How AI really works

What AI engineers do, how models read tokens, how they become assistants, why answers vary and why they make things up. Based on ideas from Chip Huyen's AI Engineering.

· What AI engineers do
· Tokens and next-word guessing
· From raw model to assistant
· Why answers change
· Why AI makes things up
· Checkpoint: How AI works

2. Measure before you trust

Why AI is hard to test, exact checks, AI judges and their biases, building a test set, and picking a model with your own tests. Based on ideas from Chip Huyen's AI Engineering.

· Why testing AI is hard
· Exact checks first
· AI as a judge
· Build your test set
· Pick the right model
· Checkpoint: Measure before you trust

3. Prompts and context, properly

Prompt anatomy, system vs user prompts, chaining and step-by-step thinking, versioning prompts like code, and defending against prompt attacks. Based on ideas from Chip Huyen's AI Engineering.

· Anatomy of a prompt
· System prompt vs user prompt
· Break it down, give it time
· Treat prompts like code
· Defend against prompt attacks
· Checkpoint: Prompts and context

4. RAG and agents

Give models the right facts with retrieval, chunk and search your notes, then let agents use tools safely, check their plans and manage memory. Based on ideas from Chip Huyen's AI Engineering.

· RAG: an open-book exam for AI
· Chunks and search
· Agents and tools
· Plans, failures and memory
· Checkpoint: RAG and agents

5. Finetuning and data

When finetuning is worth it, RAG for facts vs finetuning for form, what makes good training data, and how to use synthetic data and distillation safely. Based on ideas from Chip Huyen's AI Engineering.

· When to finetune
· Facts vs form: RAG or finetune?
· Good data beats big data
· Synthetic data and distillation
· Checkpoint: Finetuning and data

6. Ship it

Make AI apps fast and cheap, understand speed-up tricks, design the app around the model, monitor it live and learn from user feedback. Based on ideas from Chip Huyen's AI Engineering.

· Fast and cheap
· Speed-ups under the hood
· Anatomy of a real AI app
· Watch it live
· Learn from users
· Checkpoint: Ship it

Start this course free →