AI Learning Path — a 6-Week Self-Study Guide

A structured, self-paced program for anyone who wants a genuine working grasp of modern AI (the "For customer conversations" notes make it especially useful for customer-facing technical roles, but the core material is for everyone) and who wants:

• 200-level (working) understanding of the core topics: how LLMs work, prompting, RAG, running models yourself — and, above all, agentic AI, which gets three of the six weeks: the single agent, orchestration & skills, and the agent platform (control plane + data plane) that takes agents to production.
• 100-level (conversational) awareness of everything else in the landscape: fine-tuning, multimodal, reasoning models, the ecosystem — kept as a reference appendix you read opportunistically.

Time budget: ~5 hours/week for 6 weeks (~30 hours total). Each week is split roughly into 2–3 hours of reading/watching, ~2 hours of hands-on work, and 30 minutes of reflection/notes.

Prerequisites: Access to Amazon Bedrock in an AWS account (the labs use it), and the ability to run and lightly modify Python scripts. You will not need to write code from scratch — every lab is copy → run → modify. (Week 2's optional self-hosting lab additionally wants a Mac/PC that can run Ollama, and — optionally — EC2 GPU quota.)

The 6 Weeks at a Glance

Week	Theme	Depth	You'll be able to say...
1	How LLMs actually work	200	"Here's what a token is, why context windows matter, and why models hallucinate."
2	Prompting, RAG & running models yourself	200	"Here's how to get reliable output, here's RAG on a whiteboard — and here's a model I served myself with Ollama and vLLM."
3	Agents I: the single agent & agentic RAG	200	"Here's what makes something an agent, here's one I built calling real tools, and here's retrieval as a decision, not a pipeline."
4	Agents II: orchestration, skills & harnesses	200	"Here's an orchestrator delegating to skill-equipped specialists I built — and here's why Claude Code is the reference harness."
5	Agents III: the agent platform — control plane & data plane	200	"Here's the two-plane platform picture: AgentCore or EKS underneath; governance, safety, agentic evals, cost attribution, and traffic management on top."
6	Capstone & staying current	—	"Here's a working demo I built, and here's my system for keeping up."

Getting Started

1. Fork this repo (recommended) so you have your own private copy to track progress in, then clone your fork. Prefer to just read along? Clone the original directly: git clone https://github.com/aarora79/ai-learning-path.git
2. Set up the lab environment once — see labs/README.md (~10 minutes: Python, AWS credentials, Bedrock model access).
3. Open week-1-foundations/README.md and start.

The best way to use this program is to fork it, then commit as you go: check off the markdown tracker below (or the notes in your own files), and git commit after each session. Your fork becomes a dated record of what you learned and the customer-conversation takeaways you collected along the way.

How to Use This Guide

1. Do the weeks in order. Each builds on the last. If you miss a week, don't skip — shift.
2. Read the week's README first, then do readings/videos, then the lab, then write your 5-sentence summary (each week ends with a "checkpoint" — a few questions you should be able to answer out loud).
3. The labs live in labs/. Set up your environment once (10 minutes, instructions there) and every week's lab will run.
4. When something sparks curiosity, chase it — the how to find more guide shows you where to look — but come back to the path. 30 hours goes fast.
5. Keep a running doc of customer-conversation takeaways. Every week has a "For customer conversations" section. That's your real deliverable as a CSM.

Repo Map

• week-N-*/README.md — the week's agenda: an intro, a table of that week's topics, the "for customer conversations" takeaways, and the checkpoint questions
• week-N-*/*.md — one file per topic in that week (readings, key concepts, and a lab guide), linked from the week's agenda table
• labs/ — small, heavily-commented Python scripts you copy, run, and modify against Bedrock (plus one Ollama/vLLM walkthrough)
• resources/the-wider-landscape.md — 100-level reference on everything the weeks don't cover deeply: fine-tuning, reasoning models, multimodal, tokenomics, the context layer...
• resources/people-and-blogs.md — who to follow, what to subscribe to
• resources/glossary.md — plain-English definitions of every term used here
• resources/how-to-find-more.md — how to find good information on any AI topic yourself
• LICENSE — dual license: CC BY 4.0 for the content, MIT-0 for the lab code

Progress Tracker

Interactive version: open tracker.html in your browser (double-click it after cloning) — a checklist of every reading, lab, and checkpoint with progress bars and per-week notes. It saves automatically in your browser (in localStorage), so your checkboxes are not written back into the file and won't show up in git. To keep that progress in your fork or move it between machines, use the Export progress button to download a JSON file, commit it, and Import it elsewhere.

Prefer to commit as you go? The simplest durable record is the plain-markdown checklist below — edit it in your fork and git commit after each session (this is the one that actually versions your progress). Copy it into your running notes doc, or just check items off here:

• Week 1 — Foundations: readings done, lab run, checkpoint answered
• Week 2 — Prompting, RAG & local models: readings done, labs 2–4 run, checkpoint answered
• Week 3 — Agents I (single agent, agentic RAG): readings done, labs 5–6 run, checkpoint answered
• Week 4 — Agents II (orchestration, skills, harnesses): readings done, lab 7 run, checkpoint answered
• Week 5 — Agents III (the agent platform): readings done, platform whiteboard drawn, checkpoint answered
• Week 6 — Capstone: demo built and presented, staying-current system in place

License

This repo is dual-licensed, see LICENSE: the written guide (all .md files and tracker.html) is under CC BY 4.0 (reuse and adapt freely, with credit), and the lab code (labs/*.py) is under MIT-0 (use it however you like, no attribution required).

A Note on Links Going Stale

This field moves fast. Model names, product pages, and course URLs in this guide will drift. When a link 404s: the concept it teaches hasn't changed — search for the concept name plus the source (e.g., "Anthropic prompt engineering tutorial"), or check the how to find more guide.