# Track 3: AI engineering and 3D printer lab

**Final artifact:** a tested classifier or rule-based sensor concept paired with a 3D-printable object plan, print-failure log, or paper prototype. The facilitator controls printer operation and safety checks.

**Outcomes:** Students describe examples, labels, training, testing, and error; design a small physical aid; measure and revise a model; and explain why a demo is not proof of general performance.

**90-minute setup:** Prepare printed object cards, rulers, graph paper, and the [desk-organizer OpenSCAD starter](../artifacts/desk-organizer.scad). A trained facilitator checks machine operation, material, ventilation, workspace, and the model before slicing. Printing is optional.

## Sessions

1. **System and problem.** Map input, process, output, user, and human check for a classroom problem. Middle school uses cards. High school adds a failure mode and a measurable requirement. Checkpoint: system map.
2. **Data and classification.** Use [Teachable Machine](https://teachablemachine.withgoogle.com/) with non-identifying objects, or printed cards offline. Gather balanced examples, test new examples, and record mistakes. Middle school labels two classes. High school varies background and viewpoint and compares error patterns. Checkpoint: data table.
3. **Design for a physical aid.** Study a classroom need such as a cable guide, label holder, or measurement spacer. Sketch dimensions and user constraints. Middle school uses a grid and three measurements. High school adds tolerance, material choice, and a reason for orientation. Checkpoint: annotated drawing.
4. **Prototype and prepare.** Use a shared CAD or paper model. Discuss the print workflow with the [MakerBot Educator’s Guidebook](https://www.makerbot.com/educators-guidebook/). Middle school marks supports and risky edges. High school writes a slice plan and estimates material and time. Checkpoint: file or paper model plus print plan.
5. **Test and troubleshoot.** Print only after the facilitator checks the file, material, bed, and workspace. If no printer exists, inspect a provided failed print photo or simulate failure with a paper model. Record layer adhesion, dimensions, usability, and one change. Checkpoint: failure log.
6. **System demonstration.** Show the classifier and physical aid together, or present the offline equivalent. Explain where the model may fail, what the object cannot do, and what a human must check. Checkpoint: demo, test log, and maintenance note.

## Starter worksheet

| Input examples | Label | Correct? | What changed in the next test? |
|---|---|---|---|
| | | | |
| | | | |

| Object requirement | Measurement or decision |
|---|---|
| Must fit | |
| Must not do | |
| Test method | |
| Failure observed | |
| Revision | |

**Track rubric emphasis:** AI understanding, evidence and iteration, and safe team practice count twice. A strong project reports errors and measurements instead of claiming that one successful demo proves reliability.