Skip to content

Frequently Asked Questions

Answers for learners at every level. Can't find yours? Open an issue.

Getting started

Do I need a CS degree or ML background?

No. The handbook targets software engineers and career switchers who can write Python. Course 01 · Prerequisites covers the math you need. If you've never programmed, learn Python basics elsewhere first, then return to Start Here.

I'm a complete beginner. Where do I start?

  1. Prerequisites — Python + NumPy check
  2. Course 01 · GenAI Foundations
  3. Course 02 · AI Essentials — first API call
  4. Build These #1 — first portfolio piece

Skip courses 03–04 initially if you want to build first, math later.

I'm a software engineer. Can I skip foundations?

Yes — start at Course 02. Skim Course 05 for transformer context, then jump to Course 06 RAG or Course 07 Agents. See Start Here routing tables.

I'm an ML engineer. What can I skip?

Skip courses 03–04 if you know backprop and attention. Start at Course 05 LLMs, then Course 06 RAG / Course 07 Agents. Prioritize evals and production: Course 13 Evals and Course 08 Harness.

How long does the full curriculum take?

Pace Duration
Full-time 8–12 weeks
Part-time (10–15 hr/week) 4–6 months
Reference / jump-in Use Topic Map or Study Plans

Curriculum & navigation

How do I navigate the site?

Tab Use when
Start Here Pick persona, goal, prerequisites
Learn Follow courses 01–16 in order
Reference FAQ, glossary, topic lookup, resources
Projects Portfolio builds

What's the difference between Agent Engineering track and Course 07?

Resource Best for
Agent Engineering track Focused 7-lesson agent spine
Course 07 · AI Agents Full module with exercises
Course 08 · Harness MCP, permissions, runtime
Agentic AI hub Stack overview + OSS links

Pick one spine; use the others as reference.

Are exercises required?

Strongly recommended. Exercises index · Portfolio work: Build These First.

Where are capstone projects?

Technical decisions

RAG vs fine-tuning vs agents

Approach Cost Time to ship Best for Not for
Prompt engineering Lowest Hours Format, prototyping Large private knowledge bases
RAG Medium Days Q&A over your docs Changing model personality
Fine-tuning High Days–weeks Style, domain behavior Injecting fresh facts
Agents Medium–high Days–weeks Multi-step + tools Simple single-shot Q&A
Workflows Low–medium Days Deterministic pipelines Open-ended exploration 
Can prompts solve it? → Yes: prompts
                      → No: Need your docs? → Yes: RAG (Course 06)
                                              → No: Consistent style? → Fine-tune (Course 15)
                                                                      → Multi-step + tools? → Agent (Course 07)

See Course 15 · When to fine-tune · Course 07 · Workflow vs agent

RAG vs Graph RAG?

Pattern When
Vector RAG Most doc Q&A
Hybrid search Keywords + semantics matter
Graph RAG Entity relationships, "how does X connect to Y"

See Course 06 · Graph RAG lesson.

Tools & setup

What do I need installed?

  • Python 3.10+
  • pip install -r requirements.txt
  • API keys only for lessons that call live models

See Getting Started — Local setup.

Do I need a GPU?

Not for most lessons. Cloud APIs cover inference. Course 15 (fine-tuning) may use Colab — alternatives noted in lessons.

Troubleshooting

ModuleNotFoundError when running exercises

cd docs/foundations/module-01-ai-engineering-essentials/exercises
pip install openai anthropic python-dotenv
python 04-prompt-engineering-starter.py

See Exercises index.

API key errors (401)

  1. export OPENAI_API_KEY=sk-... (or .env with python-dotenv)
  2. Confirm key is active in provider dashboard
  3. Match env var name to SDK

Rate limits (429)

RAG returns irrelevant chunks

  1. Chunk size (512–1024 tokens + overlap)
  2. Embedding model match to domain
  3. Hybrid search + reranking
  4. See Course 06 · RAG evaluation

Agent loops forever

  1. max iterations in harness (Course 08)
  2. Better tool descriptions (Agent Engineering · Tools)
  3. Trajectory evals (Course 13)

mkdocs serve fails

pip install -r requirements.txt
npm run build:docs   # full build + link fix
mkdocs serve

Math overwhelming?

Career & portfolio

What should I build for a portfolio?

Three projects beat ten tutorials: RAG app, agent with tools, evaluated demo. See Build These First.

Is this enough to get a job?

Covers the technical stack. You still need shipped projects and system design practice. Use Course 16 capstones as interview stories.

How does this compare to other free resources?

We curate and extend OSS hubs with a unified path. See Open Source Hubs.

Contributing

Contribute · Roadmap · GitHub Issues