Zero to Agents
The route I'd give a new engineer on my team, stop by stop. Each stop becomes a run of short lessons: a Reel to get the idea, notes to go deeper, and code on GitHub to try it yourself.
The order isn't fixed. If a stop gets more questions, it gets more lessons. New lessons go out on the WhatsApp channel first.
- PythonFoundationsWe start here
You learn: Variables, lists, loops and functions: the parts of Python that every AI project uses.
You build: Small scripts that read a file, clean it up and answer a question about it.
- Statistics & mathsFoundations
You learn: Averages, spread, probability and just enough linear algebra to read a model.
You build: A notebook that explains a real dataset with numbers you can defend.
- Machine learning
You learn: How a model learns from examples: regression, classification, training and testing.
You build: Your first model that predicts something useful, and an honest measure of how good it is.
- Deep learning
You learn: Neural networks from the inside: layers, gradients and why depth helps.
You build: A small network trained from scratch, then the same thing in PyTorch.
- TransformersLanguage models
You learn: Attention, tokens and embeddings: the architecture behind ChatGPT and Claude.
You build: A tiny transformer that writes text, so the big ones stop being magic.
- NLP & LLMsLanguage models
You learn: Prompting, retrieval (RAG), fine-tuning and evaluating language models.
You build: A question-answering tool over your own notes.
- AI agentsAgents
You learn: Models that plan, use tools and take actions, and how to keep them reliable.
You build: An agent that finishes a real multi-step task on its own.