IntegrAuth Academy
Identity, security and practical AI — free, from zero. Byte-sized lessons, real diagrams, hands-on labs, zero fluff. Two programs, each with its own exam and certificate.
English is hard? Your browser can translate this page into Tamil, Telugu, Kannada, Malayalam or Hindi. Tech words stay in English so you learn the real names.
How?
- Chrome on phone: ⋮ menu → Translate
- Chrome on computer: right-click → Translate to …
- Edge: the translate icon in the address bar
The big picture
Ask Acme started as a short script. Now it is a full system. Your goal: follow one question from Maya through it.
- The model is the smallest part. Most fixes are in your own code.
- Everything the model reads or writes is untrusted. Put the limits outside the model.
- Measure before and after every change.
- Use the least freedom that works. Each extra step adds cost, delay and errors.
One question, start to finish
Maya types: "I paid twice by UPI for order 6021. Can you fix it?" Her browser sends only that message.
- Your first AI feature: the backend checks that she is signed in and under her quota. It adds the key and the saved chat.
- Prompts as code: each prompt is a file in git, with a version, a schema and an eval.
- Agent design patterns: a small model sends the message to the billing path. Only unusual cases go to Kai, the agent.
- Build retrieval that works: search finds the double-charge policy. The answer must cite the parts it used.
- Build an agent: no money moves before Maya says "yes". Kai gets her identity from the session. His runs have a step limit and a budget.
- Ship safely: an output guard cleans the reply and blocks outside images. The chat says it is AI and offers a person.
- Run AI in production: every step lands in a trace. A new model version goes to a few users first.
Go deeper: where these ideas go next
Much of this rests on identity. Who is the user? What may the agent do for them? The Identity & API Security program covers this:
- Governing MCP and agents
- Human-in-the-loop approvals
- Prompt injection from the identity side
- Rate limits and quotas
- Customer identity, for signing users in
In this program, look again at Prompt injection before a launch. Also read AI rules at work.
Go deeper: free courses to keep learning
These free courses were good next steps as of October 2026:
- Claude Academy: courses from Anthropic on how AI works and how to build with Claude.
- Elements of AI: a free intro for non-experts, from the University of Helsinki and MinnaLearn. It needs no programming.
- Hugging Face LLM Course: how the models work, in code, with open models.
Ready? The cheat sheet & pop quiz is next.