How KatalockAi Academy works

20 phases and 523 lessons, from the maths behind neural networks to agents running in production. You build every idea yourself before you use the tools that do it for you.

Why we built KatalockAi Academy

AI isn't just a chatbot anymore. Companies are redesigning roles around AI agents and automation, and the change is already showing up in hiring and in layoffs. The people who do well are the ones who understand how these systems work and can build with them. That's what this Academy is for.

• 116,175 US job cuts announced in 2026 where employers named AI as the reason, the most-cited reason this year. Challenger, Gray & Christmas, August 2026 report (Jan to Aug 2026).
• 170M new jobs expected worldwide by 2030 as AI and technology reshape work, while 92M are displaced. The new ones need new skills. World Economic Forum, Future of Jobs Report 2025 (January 2025).
• 59 in 100 workers will need retraining or upskilling by 2030, as nearly 40% of the skills used at work change. World Economic Forum, Future of Jobs Report 2025 (January 2025).

Governments are treating AI skills as essential, too:

• The United States has made AI literacy a national priority for students and workers, through a 2025 executive order on AI education and its AI Action Plan. GovCIO Media on the White House AI education order and AI Action Plan
• India's YUVA AI for All aims to give 1 crore people foundational AI skills. Government of India (PIB), February 2026
• Indonesia is targeting 12 million AI-skilled people by 2030. Komdigi, via Investortrust
• Singapore's SkillsFuture is giving every worker an AI readiness check, with course recommendations, from 2026. SkillsFuture Singapore, Committee of Supply 2026

Katalock is honesty-first, so every number on this page links to where it came from.

Every lesson has the same shape

1. The problem. The concrete pain the idea solves.
2. The concept. Diagrams and intuition before any maths.
3. Build it. Implement it from raw maths and plain code, no frameworks.
4. Use it. Do the same thing with production tools such as PyTorch, scikit-learn, LiteLLM, OpenTelemetry and the KatalockAi API (OpenAI-compatible), and see what each one was doing for you.
5. Ship it. Leave with something you keep: a prompt, a skill, an agent or a tool.

Learn with Professor Katalock

Every lesson has a Learn this with Professor Katalock button. It opens KatalockAi chat with Professor Katalock, the Academy's AI engineering mentor, and the full text of that lesson already loaded. The Professor teaches the way a senior engineer mentors a junior one: the intuition first, then one question to check you've got it, maths broken into small steps, one scenario question at a time, and code reviews that flag the bottleneck or memory trap before showing the clean version. In agent mode or the Code tab it can run the code and show you real output.

Lesson checks

Every lesson ends with a short lesson check: that lesson's own questions, usually five to eight. Pass it (70%) to mark the lesson as done. You can retry as often as you like, and when you're signed in your passes are saved to your account.

Projects and capstones

Every phase ends with two graded projects that make you use what the phase taught on a real task: a churn model in raw NumPy in Phase 2, a GPT you train yourself in Phase 10, a coding-agent harness in Phase 14. You build each one in your own public GitHub repository and submit it with your results. Professor Katalock reviews it against the project's published rubric (70 out of 100 to pass, no criterion below 40, copied work fails), and you can resubmit as often as you need, up to three reviews a day. Passing a phase's projects opens its final exam.

The 17 capstones are 25 to 40 hour portfolio builds, such as a real-time voice assistant, a production RAG chatbot or a multi-agent software team. Each draws on several phases, so they sit in Phases 17 and 18, and each of those phases requires one. Every capstone you pass earns its own KatalockAi certificate, linked to the repository that was reviewed.

Career routes

Career routes pick the lessons that matter for a specific AI role, such as Forward-Deployed AI Engineer, Agent Systems Engineer or AI Evaluation Engineer, in the order to take them, with a portfolio project to build along the way. Pass every lesson check on a route and its route exam, and you get a career route certificate on top of your phase certificates. Six shorter skill paths (5 to 17 lessons, such as Model Context Protocol or Agent Skills Engineering) end in a 10-question path quiz and a skill path mini certificate.

Prefer your own coding agent? (optional)

The same tutor is available as a portable skill for agent hosts that read SKILL.md files: Claude Code, Codex, Cursor or KatalockAi itself.

• KatalockAi: open Customize, then Skills, and import this link: katalock.net/skills/katalock-academy-tutor/SKILL.md
• Claude Code, Codex or Cursor: download SKILL.md and save it as katalock-academy-tutor/SKILL.md inside your agent's skills folder (for example ~/.claude/skills/ or ~/.cursor/skills/).

Final exams and certificates

Each phase ends with a final exam, which opens once you've passed the phase's projects: 20 questions drawn at random, 30 minutes, 70% to pass, up to 3 attempts a day. Pass it and you get a KatalockAi certificate in your name with its own certificate ID and the full list of topics you covered. Anyone, including an employer, can check that ID at katalock.net/academy/verify.

Certification prep

Preparing for an industry exam? Certification prep covers Anthropic's four Claude certifications and the Model Context Protocol Associate (MCPA) exam, with lessons, labs, study plans, a free diagnostic and full-length practice exams for each track. It's independent study material: it isn't affiliated with the exam owners and Katalock doesn't issue certificates for it.

What's free

Phases 0 to 3 are free, with their projects, exams and certificates, plus the first lesson of every other phase. Any paid KatalockAi plan unlocks every phase.