AI Under the Hood — Experiential AI Workshop
What it is
AI Under the Hood is a live, in-person workshop that shows how a trustworthy AI answer is actually built, using the working STEMVentor Mentor as the example from start to finish.
What makes it different
Most AI in education is a chatbot added to a course catalogue. It does not know what the learner is looking at, and it still guesses. The Mentor was built differently: it is grounded in a knowledge graph, so a question is answered by finding the right material, checking what is related to it, and only then responding. The workshop walks participants through that design step by step and live, so they leave understanding how such a system works and how to judge one.
What it includes: the journey
The session follows fourteen steps, from the basics of AI to a retrieval pipeline running in front of the room. The content draws on two of the AI learning paths, The AI Landscape and Build Your Own AI.
| Step | Topic | What participants learn |
|---|---|---|
| 1 | Welcome | Where the session is headed |
| 2 | Human vs. AI, machine learning types, capability levels | What makes something artificial intelligence, and why AI is not one thing but a range of abilities |
| 3 | Industry applications | Where AI already works quietly in everyday life, across industries they interact with |
| 4 | Algorithms-as-a-Service | Ready-made AI services that companies plug into their products instead of building from scratch |
| 5 | Generative AI and prompt engineering | How tools like ChatGPT generate answers, and how the way a question is asked changes what comes back |
| 6 | Agentic AI: tool use and reasoning loops | How AI moves beyond answering, taking a goal and working through several steps on its own |
| 7 | LLMs and talking to one through an API | What happens behind the scenes between typing a question and receiving a reply |
| 8 | Embeddings | How a computer learns what text means, not only which words match |
| 9 | Knowledge graphs | How connecting related content, like a mind map, gives AI a sense of context |
| 10 | Graph-augmented RAG | Shown live: finding the right information, checking what is related, and only then responding |
| 11 | Session awareness and personalization | How an assistant can tell what a user is working on and tailor its help |
| 12 | Grounding, guardrails and offline-first | How to tell an honest AI from one that makes things up, and why AI should never be the one thing a product cannot work without |
| 13 | Open floor and live demo | Participants ask the live system anything and see why it answers the way it does (full-day only) |
| 14 | Closing | How the same approach applies beyond this one platform |
Formats
The half-day session runs four hours and covers the complete theory with one live demo. The full-day session runs eight hours and covers the same theory with two live demos, an architecture walkthrough and the open-floor session. Group size is not limited.
How an institution can use it
The workshop is offered to schools and educational organizations, and to businesses and organizations that want their people to understand AI at this level. It runs as a Learner session or as an Educator and Trainer session. The content is the same, and the framing adapts to the room.
What participants leave with
- A working vocabulary of AI types and capability levels, and a picture of where AI is already applied.
- A way to decide when to use a ready-made AI service and when to build.
- An understanding of how a language model is called through an API, how embeddings capture meaning, and how a knowledge graph adds context.
- A view of retrieval-augmented generation as it happens, not only as a diagram.
- A way to judge whether an AI system is grounded or guessing, and why offline-first design matters.
An introductory session, Getting Started with AI, is available for audiences new to the subject. Details of the workshop are at stemventor.com/ai-under-the-hood.