Free · Project-first · Beginner → Agentic
From "Java dev who knows no Python" to RAG, LLMs and agentic systems — through explainers you drive, engines you build from scratch, and systems you ship to production.
Ten phases, each with topic-by-topic guides, checkpoints, and a portfolio artifact. Click a phase for its full guide.
The Path is the answer to "what do I do next"; everything below is a tool you'll use along it.
🧭Start here — one item at a time, in order. Your ordered queue through explainers, quizzes, exercises and projects, with progress kept on this device.
Start here →262 concepts you can click, drag and break — backprop, attention, RAG, agents. The same explainers The Path steps you through, from Phase 0 on.
Start seeing → 🤖The game: earn GRAD-1's six parts in mini-games — leakage, splits, losses, learning rates. A game for classic ML — play it when The Path reaches Phase 3.
Start forging → 🔨16 staged, test-driven from-scratch builds: autograd, mini-PyTorch, GPT, RAG engine, vector DB. Keyed to the path — BYO-1 unlocks at Phase 0, BYO-13 at Phase 8.
Start building → 🏗️8 deployable systems around the model — RAG with verified citations, guardrails, agent sandboxes, evals. The core 8 start once The Path reaches Phase 6.
Start shipping → 🏛️The sibling site: CCAR-F exam prep as a game — the Agent Hangar, 26 explainers, quizzes, missions. Not part of this path.
Enter the hangar → 💼What top companies actually test in 2026, a role map for AI Engineer/FDE/AI Architect/ML Engineer/Research Engineer, and a readiness tracker built from your own progress here.
See the role map →