Designing a Curriculum Unit on Generative AI for High School CS Using Raspberry Pi HATs

Hook: Solve the “too many tools, too little time” problem with a classroom-ready unit that actually builds hireable skills

Teachers: you face a familiar set of frustrations — students excited about AI but confused by jargon, limited class time, and endless online courses that don’t translate into portfolios. This unit uses the Raspberry Pi 5 paired with the AI HAT+2 to deliver structured, hands-on lessons in generative AI that produce portfolio-ready projects, measurable assessments, and guided ethics conversations suitable for high-school computer science classes in 2026.

Why this matters in 2026: on-device generative AI is classroom-ready

Two trends made this unit possible and urgent in 2026: (1) the rise of efficient, high-quality models optimized for edge devices and (2) mainstream vendor moves to integrate large multimodal models into everyday devices. Recent coverage in late 2025 highlighted the release of the AI HAT+2, which "unlocks generative AI for the Raspberry Pi 5" and brings on-device inference into reach for classrooms. Meanwhile, industry partnerships—like major voice-assistant integrations using Gemini-class models—have underscored how generative AI is moving from cloud-only demos to hybrid and on-device deployments.

"The new $130 AI HAT+2 unlocks generative AI for the Raspberry Pi 5." — ZDNET, 2025

That means students can experiment with text generation, simple multimodal outputs (images + text), and small local LLMs without sending their data to external servers — a huge win for privacy, cost, and classroom scalability.

Unit at a glance (4 lessons + capstone)

• Grade level: 10–12 (adaptable)
• Duration: 4 class periods (60–90 min) + 2–4 weeks capstone
• Tech: Raspberry Pi 5, AI HAT+2, microSD, power supplies, optional camera module/mic, teacher workstation
• Learning goals:
  • Understand core concepts of generative AI and on-device inference
  • Build and evaluate a simple generative demo (text, image, or multimodal)
  • Apply ethics frameworks to assess model outputs and biases
  • Create a portfolio artifact and reflect on technical and soft skills

Standards alignment & competencies

This unit maps to common CS standards and career-readiness competencies: computational thinking, data privacy, model evaluation, project planning, collaboration, and communication. It supports CSTA standards for algorithms and AI literacy, and contributes to employability skills such as problem-solving, documentation, and ethical reasoning.

Materials & prep (teacher-friendly checklist)

• One Raspberry Pi 5 per group (2–3 students per Pi recommended)
• AI HAT+2 accessory and mounting hardware
• microSD cards (32GB+), power supplies, HDMI or headless access
• Optional: camera module, USB microphone, small speakers
• Teacher workstation with flashing software and network access (consider recommended ultraportables if you're buying hardware)
• Preloaded image: Raspberry Pi OS + AI HAT+2 SDK (create a master image to save class time)
• Printed rubrics, handouts, and ethics prompt cards

High-level setup (15–30 minutes prep per Pi)

1. Flash Raspberry Pi OS image and install latest AI HAT+2 drivers/SDK. Use a single master image and clone to microSDs to streamline class setup.
2. Connect the AI HAT+2 and confirm device recognized (basic shell command tests or the SDK’s diagnostic tool).
3. Install lightweight inference frameworks (PyTorch/XLA or vendor SDK) and example notebooks for text generation and image generation tasks.
4. Verify offline demo runs at tolerable latency — if network is limited, confirm all required assets are local.

Lesson 1: Generative AI Fundamentals (60–75 min)

Learning objectives

• Explain what generative AI is, and the difference between cloud and on-device models
• Describe basic model behaviors: creativity, repetition, hallucination
• Run a teacher-led demo of a local text-generation model on the AI HAT+2

Activities

1. Hook (10 min): Show a quick on-device demo — prompt a local model to write a short poem about the school mascot. Ask: "Would you trust this for a research report? Why or why not?"
2. Mini-lesson (15 min): Present core concepts: tokens, prompt, parameter scale, latency, privacy trade-offs for on-device vs cloud.
3. Guided lab (25–40 min): Students pair up. They run a notebook that loads a small language model on the AI HAT+2 and experiment with prompts to observe outputs. Provide starter prompts and a short worksheet to capture observations (tone, accuracy, hallucination instances).

Deliverable

Worksheet with 3 sample prompts, observed outputs, and a short reflection on one hallucination or factual error found.

Lesson 2: Multimodal & Creative Demos (75 min)

Learning objectives

• Understand basic image-generation and multimodal prompts
• Build a simple text-to-image or captioning demo using the AI HAT+2
• Start a small team project idea for the capstone

Activities

1. Demo (10 min): Teacher shows a text-to-image generation on-device (pre-generated sample if live latency is high).
2. Lab (45–50 min): Students load an example script that either:
   • Generates a small 256×256 image from a prompt, or
   • Captions a supplied image from the Pi camera
3. Reflection (10–15 min): Use a guided rubric to critique outputs for creativity, fidelity to prompt, and potential ethical issues (e.g., stereotypes in generated images).

Deliverable

Screenshot or saved image with the prompt and a one-paragraph analysis.

Lesson 3: Evaluation, Robustness & Safety (60–75 min)

Learning objectives

• Assess model output quality with quantitative and qualitative metrics
• Identify bias, hallucination, and privacy risks
• Learn mitigation techniques for on-device deployments

Activities

1. Quick lecture (10 min): Introduce evaluation criteria: relevance, accuracy, fluency, bias detection, and robustness to adversarial prompts.
2. Group exercise (25–30 min): Each group runs the same set of prompts and completes an evaluation matrix. Compare across groups to see variability and discuss causes (prompt phrasing, model randomness, temperature).
3. Ethics workshop (20–30 min): Use role-play cards to debate scenarios — e.g., using AI-generated content for a school publication without disclosure; training a model on student data. Provide a short checklist for responsible use.

Deliverable

Evaluation matrix and a short group statement of recommended safeguards for their capstone.

Lesson 4: From Prototype to Portfolio (60 min)

Learning objectives

• Plan a capstone project and map deliverables to rubric criteria
• Write clear documentation and a short project narrative suitable for a resume
• Practice presenting technical work to a nontechnical audience

Activities

1. Capstone planning (25 min): Students form teams and choose a project (see suggested capstones below). They draft an implementation plan with milestones.
2. Documentation workshop (20 min): Teach a one-page project summary template: Problem, Approach, Model/Tooling, Results, Ethics/Limitations, What I learned.
3. Lightning presentations (15 min): 2–3 teams present their plan and receive peer feedback focused on feasibility and ethical considerations.

Deliverable

Capstone project plan and a draft one-page project summary ready for portfolio inclusion.

Capstone project ideas (portfolio-ready)

• Local study assistant: On-device Q&A for a specific topic (e.g., biology chapter), with citation-checking heuristics.
• School newsletter writer: A generative tool that drafts event summaries, with a human-in-the-loop approval workflow and bias checks.
• Accessibility helper: Captioning and simple image descriptions for school photos, focused on accuracy and privacy.
• Art generator for school yearbook: Controlled prompts and style transfer with explicit copyright-safe source handling.

Assessment: Rubrics and scoring (teacher-ready)

Use a combined rubric that measures technical skill, communication, teamwork, and ethical reasoning. Suggested weight: Technical 40%, Communication & Documentation 25%, Ethics & Safety 20%, Collaboration 15%.

Rubric template (4-level scale)

• Exceeds Expectations (4): Work demonstrates advanced technical understanding, reproducible demos, clear documentation, and active ethical mitigation.
• Meets Expectations (3): Functional prototype, adequate documentation, and clear consideration of safety/ethics.
• Developing (2): Prototype runs but lacks robustness or clear documentation; ethics discussion superficial.
• Beginning (1): Incomplete or nonfunctional demo; missing documentation and ethics reflection.

Rubric criteria (sample descriptors)

• Technical correctness: Model loads and runs, meets latency/accuracy targets, code organized and commented.
• Experimentation & results: Clear test cases, evaluation matrix, and comparative analysis of outputs.
• Documentation & presentation: One-page summary, README, and a 3–5 min demo presentation targeted at nontechnical reviewers.
• Ethics & safety: Identifies specific risks (bias, privacy), applies at least two mitigation strategies, and documents limitations.
• Collaboration: Roles assigned, contributions tracked, and peer reviews completed.

Classroom management & teacher hacks

• Use a master image and clone microSDs to save setup time. Pre-run heavy installs in a lab image.
• Group size 2–3 per Pi keeps engagement high and reduces hardware bottlenecks.
• Keep network access optional: design for offline use so you avoid bandwidth or firewall issues during demos. See guidance on power resilience and low-budget retrofits for makerspaces and classroom labs.
• Use checklists and a quick troubleshooting document for common issues (drivers, permissions, overheating during long runs).
• Encourage students to maintain a simple Git repository for project code and documentation — teaches version control basics.

Safety, privacy & ethics discussion prompts

Ethics and safety are not optional. In 2026, with more on-device capabilities, students must learn not only how to build but how to judge responsible use.

• Prompt: "If a local model hallucinates a quote and a student posts it to the school website, who is responsible? How would you prevent this?"
• Prompt: "We trained on images scraped from the web; what are the copyright and consent concerns?"
• Prompt: "How could your tool affect marginalized groups? Run a quick bias audit on sample prompts and report findings."
• Prompt: "What data would you collect from users of your capstone app, and how would you minimize risk?"

Assessment: Give students a short reflective rubric item — describe two risks and one technical mitigation; grade on completeness and realism.

Sample teacher demo script (concise)

1. Boot Raspberry Pi 5 with AI HAT+2 and open a prepared Jupyter notebook.
2. Load the small on-device model, demonstrate prompt-to-text generation (keep examples PG).
3. Show an example of a hallucination and discuss how you verified it was wrong.
4. Switch to a text-to-image cell or captioning demo. Pause to discuss copyright/sensitive content rules before generation.

Resources & further reading (curated for teachers)

• Vendor docs and SDK for AI HAT+2 (recommended: pre-download and archive for offline classroom use)
• Short primers on model evaluation and bias audits (select classroom-friendly guides from 2024–2026)
• Sample GitHub classroom repositories with starter notebooks (create a private org for student repos)
• District/legal guidance on student data privacy — consult before collecting any personally identifiable information

Extensions for advanced students

• Optimize model latency and memory: experiment with quantization and pruning techniques suitable for the AI HAT+2.
• Build a simple human-in-the-loop pipeline: automatic generation + teacher approval UI.
• Explore multimodal chaining: connect short text output to an image generator and then to a classifier for content filters.

Real-world classroom case (experience & evidence)

In late 2025 and early 2026 pilot programs used Raspberry Pi 5 + AI HAT+2 to teach applied AI in small groups. Teachers reported higher engagement and clearer portfolio artifacts compared with cloud-only demos because students owned the full stack from prompt to output. Portfolios from these pilots were used successfully in internship applications and community showcases — concrete evidence that on-device projects translate to job-readiness.

Common challenges and troubleshooting

• Performance trade-offs: small models have limits — set expectations about fidelity and latency early.
• Installing dependencies: prebuild images to avoid hours of per-device installs.
• Content safety: implement simple filters and a human approval step before publishing student-generated content.
• Assessment fairness: use checklists and peer review to ensure consistent grading across groups.

Future-proofing: why this unit stays relevant beyond 2026

On-device generative AI will continue to grow as models become more efficient and hybrid deployments more common. Teaching students how to evaluate, deploy, and ethically manage generative systems on commodity hardware gives them transferable skills for internships and entry-level roles: reproducible engineering, prompt and prompt-engineering literacy, evaluation metrics, and clear ethics reasoning.

Actionable takeaways (what to implement this week)

1. Order or secure one Raspberry Pi 5 and AI HAT+2 for every 2–3 students and create a master microSD image.
2. Prepare one working demo notebook (text generation) and one ethics prompt pack for day 1.
3. Use the provided rubric weights (Technical 40%, Ethics 20%, Comm 25%, Collaboration 15%) as your baseline grading policy.
4. Plan a capstone that produces a one-page portfolio artifact and a 3–5 minute demo — that’s what internship reviewers want.

Closing: equip students for real work with small, meaningful projects

Generative AI education in 2026 is most effective when it's hands-on, ethical, and portfolio-focused. Using the Raspberry Pi 5 with the AI HAT+2 lets students take ownership of the entire pipeline — from code to consequences — and graduate with demonstrable skills employers recognize. With clear rubrics, scaffolded lessons, and an emphasis on ethics, you can run this unit in a single term and leave students with concrete work for resumes and interviews.

Call to action

Ready to run this unit next term? Download the teacher pack (lesson slides, starter notebooks, printable rubrics, and ethics cards) and join our educator community forum to share capstone projects and assessment examples. Turn curiosity into career-ready skills — start your pilot this semester.

Related Reading

• Benchmarking the AI HAT+2: Real-World Performance for Generative Tasks on Raspberry Pi 5
• Low‑Budget Retrofits & Power Resilience for Community Makerspaces (2026)
• Review: Best Sticker Printers for Classroom Rewards (2026)
• Beyond Filing: The 2026 Playbook for Collaborative File Tagging, Edge Indexing, and Privacy‑First Sharing
• Campus to Career 2026: Micro‑Credentials, Short‑Form Assessment, and the New Apprenticeship
• Protecting Your Crypto and Tax Refund from Government Offset: What Crypto Traders Need to Know
• From Studio to Screen: Careers in Scoring Big Franchises After Hans Zimmer’s Move to Harry Potter
• Inside JioStar’s Profit Engine: What the INR8,010 Crore Quarter Says About Indian Media Economics
• Streaming Mexican Food: Lessons from JioStar’s Big Play for Food Media Creators