Python 3
Advanced IntermediateModules, files, errors, clean notebooks. Not production microservices.
Ages 15–18
AI Engineers
Ages 15–18 treat AI as engineering — measure errors, state limits honestly, and ship a demo a parent can try.
Class
1 : 4 · 4 children
Term
24 studios · 12 weeks
Weekend
Friday + Saturday
Length
75 minutes each
Who it is for
15–18s who want AI they can actually run and explain. A short diagnostic first. Gaps in Python are filled or we place into App Builders.
Not a mystery box. They train a small model, build a scoped vision demo, and finish a capstone with a system card — Friday and Saturday labs, four engineers, one mentor.
How the term runs
Weekend studios, two days a week.
Fri & Sat · 18:45–20:00 · AST / GST · 75 minutes
60–90 minutes between studios · English · live online.
Friday
Lab: train, try, break, note what happened.
Saturday
Critique: where did that answer come from?
Ship
A parent can type, point a camera, or read the card.
Software they grow confident in
Not a logo list. Each line is the level they reach by the end of the term — and what that means in practice.
Jupyter / VS Code notebooks
IntermediateReproducible runs, clear variables, and plots they can explain.
NumPy · pandas · matplotlib
IntermediateTables, splits, charts — say what a column means.
scikit-learn
Basic → IntermediateTrain/test split, a classifier, metrics, confusion matrix.
OpenCV (scoped) · Streamlit / Gradio
BasicImage as data; a safe demo UI with stated limits.
GitHub
IntermediateRepo hygiene, README, and a publishable model/system card.
Three projects this term
At each stage the mentor helps them choose. By showcase night, three builds exist that the family can open.
Project 1
Studios 1–8
Spam/filter, simple predictor, or Data Detective
They ship: Notebook + small demo with data sheet and metrics
Project 2
Studios 9–12
Object Lens (vision) or a scoped gesture/posture demo
They ship: Safe demo + wrong-prediction gallery + intended-use note
Project 3 · Capstone
Studios 13–24
Notes chatbot, constrained agent, or Color Rover sim with logs
They ship: Repository, demo/walkthrough, system/model card, defence
Term shape
How the twelve weeks move — foundations, builds, polish, showcase. Lesson-by-lesson teaching plans stay with the mentor.
- 1
Data & honest models
Studios 1–8AI as engineering, data quality, baselines, metrics, bias — then Project 1 ships.
- 2
Vision with limits
Studios 9–12Pixels to predictions, error galleries, and a demo that states what it must not be used for.
- 3
Systems & agents
Studios 13–18Sensing-to-action or tool boundaries, logging, and red-team evaluation.
- 4
Capstone & defence
Studios 19–24Proposal, build, measure, write the card, rehearse, and showcase.
By the end of the term
What they can do on their own
Cold-start the project, show a metric, explain a wrong prediction, state intended use and refusal cases, and defend the work in twelve minutes.
- A responsible classifier with data sheet and metrics
- A scoped vision demo (Object Lens) with an error gallery
- A capstone system + model/system card and a 12-minute defence
Inside the studio
AI as engineering — measurable, auditable, explainable.
Studio format
Studio size
1 mentor · 4 engineers · seminar + lab in the same studio
Cadence
Friday + Saturday · 24 studios over 12 weeks · GST / AST
Language
English · technical writing parents and universities can read
Standard
If they cannot restart it and explain a failure, it is not finished
Craft in the room
Data & models
Clean a small dataset, train or fine-tune lightly, measure — no mystery “AI magic” buttons.
Vision & perception
Cameras, classifiers, and confidence they can plot — when the model is wrong, they say so.
Agents & tools
Pipelines that call tools, log steps, and stop safely — built to be audited, not demoed once.
Systems discipline
README, environment, logs, and a cold start. University-ready hygiene, not slide decks.
Mentors demand the why. Hype slides fail. A calm walkthrough of data, model, and failure modes is the only pass.
Project briefs
Open the live demos below. These are the kinds of builds in AI Engineers — games, sites, and tools a family can try on the call.
AI chatbot
Notes Bot
Answers only from a notes pack — refuses when outside scope.
Python logicRetrievalRefusal
Number reader
Digit Spot
Draw a digit; get a guess and a confidence number.
CanvasFeaturesConfidence
AI chatbot
Oasis Bot
Chat grounded in study notes — refuse homework it should not do.
PythonChatNotes search
Camera AI
Object Lens
Point a camera, get a label and confidence — admit when unsure.
PythonVisionLabels
Ready when you are
Let them learn to code — for real.
We’ll check AI Engineers is the right fit for their age and what they like to make.