calmrocks/ai-engineer-notebooks

What it claims

Beyond what the summary states, the page pins down the exact curriculum order and two sections the summary omits: 10 — ML system design & performance and 11 — Customer craft (the FDE differentiator), alongside 00 Setup, 01 Model APIs, 02 Evals I, 03 RAG, 04 Evals II, 05 Agents, 06 Adapting the model, 07 Security, 08 Operations, 09 Serving & inference performance, and 12 Case Studies & Capstone. The three case studies are named concretely: a support assistant debugged in production, a pipeline-vs-agent cost showdown, and a red-team robustness benchmark.

How to use it

Everything runs on the free Groq API (no credit card). Notebooks are self-contained: the first cell installs dependencies, the second calls the repo's tiny shared-setup package to load your key:

from aien import setup
client, MODEL = setup()

In Colab, add GROQ_API_KEY as a secret via the key icon and toggle notebook access on. To run locally instead:

pip install -r requirements.txt && pip install -e .
export GROQ_API_KEY=...

(the pip install -e . step installs the aien helper), then open with Jupyter.

Portability is explicit: the notebooks are OpenAI-compatible throughout, so patterns transfer directly to OpenAI and, with small changes, to Anthropic — swap the base URL. A shared corpus in data/ is reused across the RAG and eval sections so evals measure the retrieval you actually built.

The capstone is positioned separately from the case studies: case studies are runnable learning scenarios, while the capstone is a brief for a deployed repo you build yourself with a serving component and an eval report, intended for a resume.

Limits

Two topics cannot run on Groq because they need a GPU: LoRA fine-tuning (06) and self-hosted serving (09). These are taught concept-first with optional, fenced Colab-GPU appendices verified on a real Colab T4 — so hands-on work there requires GPU access the free tier does not provide. The agents section's coverage is stated to include tool design, guardrails, MCP, and Skills; adaptation is framed as fine-tuning vs LoRA.