AI Engineering from Scratch: choose a route through the curriculum
Set up a study environment without turning a course into a service
Separate local exercises, optional tutor installation and the Docker lesson.
What you will learn
- Choose the environment per lesson
- Bound the optional agent skill install
- Keep credentials and data outside exercises
Before you start
- Basic command-line access
- A bounded learning goal and disposable checkout
Turn lesson commands and code changes into a reviewable record of what you can do.
Key takeaways
- A Docker lesson is not a deployment of the whole course.
- The tutor is optional and changes an agent host when installed.
- Local learning exercises should avoid live customer data.
Choose the environment per lesson
The course includes Python, TypeScript, Rust and Julia examples, but a learner does not need every runtime on day one. Install the language and dependencies required by the first route, then add tools as a lesson calls for them.
The setup phase has a Docker-for-AI lesson with its own Dockerfile and Compose example. That material teaches container concepts; it is not a deployment manifest for the entire curriculum. Treat it as an isolated exercise.
Bound the optional agent skill install
The README offers `npx skills add rohitg00/ai-engineering-from-scratch` for a placement tutor and learning skills. That installer can write into a chosen agent host and scope. Inspect the destination, installed skill files and trust prompts before enabling them on a work machine.
If Node or a supported host is unavailable, use the website or repository lesson files. Only agent-host discovery and invocation evidence remains untested on that route; the core written material does not disappear.
Keep credentials and data outside exercises
Later phases discuss APIs, notebooks and datasets. Use a separate environment for practice, keep provider credentials in the platform’s secret mechanism and start with public or synthetic data.
A study setup is successful when a chosen lesson runs and its output is explainable. No production endpoint, autoscaling plan or service-level target is implied by the course repository.
Decision guide
| Criterion | Option A | Option B |
|---|---|---|
| Best when | You need predictable behavior and easy auditing | You need adaptive optimization and have reliable telemetry |
| Main risk | May leave performance on the table | Can become difficult to explain or debug |
Implementation steps
- 1
Install only the runtime needed for your first route.
- 2
Inspect the optional skill installer’s target scope.
- 3
Run the Docker lesson in an isolated environment if it belongs to your route.
Copy-ready example
git clone https://github.com/rohitg00/ai-engineering-from-scratch.git
cd ai-engineering-from-scratch
python3 phases/00-setup-and-tooling/01-dev-environment/code/verify.py --route beginnerFrequently asked questions
Can I deploy this repository as an AI product?
It is a curriculum and set of exercises; deploy a specific project you build from it, with its own design and checks.
Do I need a GPU for the first lesson?
The documented beginner preflight and linear algebra example can start without one.
Sources
- AI Engineering from Scratch / README.mdSource checked 2026-09-29
- AI Engineering from Scratch / phases/00-setup-and-tooling/07-docker-for-ai/docs/en.mdSource checked 2026-09-29