The AI Path

What it is

After three years as a teaching assistant for Machine Learning, Data Mining and Algorithm Design, I now teach in the open. The AI Path is a free course with one promise: learn AI by building it, together. It is paced for people who can code a little, in any language, but for whom Python and AI still feel like a wall. There is a separate starting point for people who have never coded, and a skip guide for people who already work in tech.

We learn it. We build it. We break it. And we keep moving forward.

How it teaches

Python is the tool, not the goal. Each episode teaches one core idea that holds in any language, builds it step by step, and shows where it lives beyond Python, so that when the tools change the learner does not start over.

Every build follows the same rhythm:

PlanPredictRunCheckExplain

A week has three parts. Watch the episode: one idea, one build, and one bug broken on purpose and then fixed. Open the notebook in Colab, free and in the browser. Do one mission, at whichever level fits the week:

• Starter, about ten minutes: run the notebook, change one thing, share the result.
• Core, 30–45 minutes: the real coding task, saved to the learner’s own fork.
• Stretch, 60–90 minutes: a portfolio extension, or a “rebuild it another way” challenge.

Solutions appear in each episode’s folder seven days after the episode. Try first.

The map

1. 1

   Season 1: Foundations

   Python for AI, real data, SQL, statistics by simulation

   Project An end-to-end data and SQL investigation

2. 2

   Season 2: Machine learning

   Honest evaluation first, then the classic models

   Project A machine learning pipeline you can trust, with a model card

3. 3

   Season 3: Deep learning

   Neural networks from scratch, computer vision, a tiny Transformer

   Project A computer-vision project and a Transformer you wrote yourself

4. 4

   Season 4: Large language models

   Tokens, embeddings, RAG and evaluation

   Project An AI app you can evaluate, not just demo

5. 5

   Season 5: AI engineering

   Tool-using agents, deployment, monitoring, cost, security

   Project A deployed AI agent with guardrails

6. 6

   Season 6: Career and research

   Portfolio reviews, interview prep, reading papers

   Project Your portfolio, reviewed

Season 1, Foundations, runs from the map of the course to an end-to-end data investigation: a first model in Colab, Python for AI, vectors and NumPy, cleaning real data with pandas, exploratory analysis, SQL, statistics learned by simulation, and gradient descent written from scratch.

The rules I hold it to

• Every notebook runs top to bottom from a fresh runtime, with pinned versions.
• No fake outputs, made-up numbers or staged results.
• When a tool changes or I make a mistake, I fix it in the repository with a commit, and with a pinned comment under the video.
• I ask permission before showing anyone’s work or comment in a video.