⚡ The Rural Connectivity Challenge

Why Cloud-Only AI Fails Rural Schools

Malaysian educators in Sabah, Sarawak, and remote peninsular regions face a daily reality that city-based teachers rarely encounter: unstable or absent internet.

  • 📡 Latency & dropped connections

    Cloud AI tools like ChatGPT and Gemini require stable broadband. A lesson interrupted by a timeout wastes precious class time.

  • 💰 Mobile data costs

    Many rural teachers use mobile hotspots. Streaming large AI model responses consumes expensive data plans.

  • 🔒 Privacy & data sovereignty

    Student data should stay local. Offline tools keep sensitive classroom data on your own device.

The solution: Run AI models locally on your laptop or desktop using Ollama — no internet needed after the initial one-time setup.

đŸĻ™ Ollama — Fully Offline Local AI

What is Ollama?

✓ Works Offline

Ollama is a free, open-source tool that runs large language models entirely on your computer. Once you download a model (one-time, ~2–4 GB), it runs with zero internet connection — even in the deepest jungle.

System requirements: Works on Windows, Mac, and Linux. Runs acceptably on 8 GB RAM school laptops.

Download Ollama (free) →

Recommended Models for Malaysian Schools

These models are small enough to run on school hardware and strong enough for classroom tasks:

Offline
gemma2:2b
by Google ¡ ~1.6 GB

Low resource usage, strong reasoning. Great for lesson planning and Q&A generation.

Offline
phi3:mini
by Microsoft ¡ ~2.2 GB

Excellent for structured tasks like lesson plans, rubrics, and formatted reports.

Offline
qwen2.5:3b
by Alibaba ¡ ~2.0 GB

Multilingual (Malay & Chinese support). Ideal for Malaysian classrooms with diverse language needs.

Offline
llama3.2:3b
by Meta ¡ ~2.0 GB

General-purpose, strong in English. Good for feedback, translations, and explanations.

# Install and run (one command)

ollama run qwen2.5:3b

What You Can Do Offline

📝

Generate Lesson Plans

Describe your topic, year level, and curriculum standard — get a full lesson plan in seconds.

✅

Grade & Give Feedback

Paste student essays or answers — get constructive feedback without sending data to the cloud.

🌐

Translate Materials

Translate between English, Bahasa Malaysia, and Chinese using qwen2.5 for multilingual classrooms.

⚡ Claude Code CLI Skills

Claude Code's skills system lets educators invoke pre-built AI workflows from the command line — like calling a specialist by name. Each skill is a ready-made prompt workflow tailored to a specific educational task.

Use these skills with Claude Code CLI when you have internet, or paste the skill descriptions from docs/skills.md into Ollama when you don't.

Teaching & Curriculum Design

5 Skills

Lesson Plan Generator

æ•™æĄˆį”ĸį”ŸåˇĨå…ˇ

AI-assisted lesson plan generation aligned with KSSR/KSSM curriculum

teaching-design-lesson-plan-generator

Adaptive Learning Design

č‡ĒéŠæ‡‰æ•™å­¸č¨­č¨ˆ

Design adaptive learning pathways tailored to individual student needs

teaching-design-adaptive-learning-design

Differentiated Instruction

åˇŽį•°åŒ–æ•™å­¸

Create materials for mixed-ability classrooms

teaching-design-differentiated-instruction

Project-Based Learning

å°ˆéĄŒåŧå­¸įŋ’

Design PBL units with real-world connections

teaching-design-project-based-learning

Competency-Based Learning

æ ¸åŋƒį´ é¤Šæ•™å­¸

Develop 21st-century competency-based learning activities

teaching-design-competency-based-learning

Learning Analytics

4 Skills

Learning Analytics

å­¸įŋ’分析

Analyze student data to identify patterns and improve outcomes

learning-analytics-learning-analytics

Competency-Based Assessment

į´ é¤Šå°Žå‘čŠ•é‡

Design assessments aligned with learning outcomes

learning-analytics-competency-based-assessment

Peer Assessment

åŒå„•čŠ•é‡

Design structured peer assessment activities and rubrics

learning-analytics-peer-assessment

Student Profiling

å­¸į”Ÿį•Ģ像分析

Create comprehensive student learning profiles

learning-analytics-student-profiling

Research Methods

4 Skills

Educational Research Design

æ•™č‚˛į ”įŠļ設計

Design educational research studies with appropriate methodology

research-methods-education-research-design

Empirical Research Automation

å¯Ļ證᠔įŠļč‡Ē動化

Automate data collection and analysis for empirical research

research-methods-empirical-research-automation

Mixed Methods Research

æˇˇåˆæ–šæŗ•

Combine quantitative and qualitative research approaches

research-methods-mixed-methods

Quasi-Experimental Design

æē–å¯Ļ驗設計

Design quasi-experimental studies for educational interventions

research-methods-quasi-experimental-design

Rural & Inclusive Education

2 Skills

Localized Content Design

åœ¨åœ°åŒ–å…§åŽšč¨­č¨ˆ

Design locally relevant content for rural Malaysian contexts (Sabah, Sarawak, Orang Asli)

rural-education-localized-content-design

Rural Education Integration

åé æ•™č‚˛æ•´åˆ

Integrate technology and community resources for rural schools

rural-education-rural-education-integration

AI & Educational Technology

2 Skills

AI Teaching Tools

AI 教學åˇĨå…ˇ

Evaluate and implement AI tools appropriate for Malaysian classrooms

ai-tools-teaching-tools

Generative AI Education

į”Ÿæˆåŧ AI æ•™č‚˛

Design curricula and activities teaching responsible AI use

ai-tools-generative-ai-education

ICT, Technology & Mathematics

3 Skills

ICT Integration

ICT æ•´åˆčžå…Ĩ

Integrate ICT tools across subjects following MOE Malaysia guidelines

ict-ict-integration

Math-Technology Education

æ•¸å­¸į§‘æŠ€æ•™č‚˛

Use GeoGebra, Desmos, and other tools to enhance Mathematics teaching

math-tech-math-technology

📄 The skills.md Reference File

What is skills.md?

docs/skills.md is a plain-text reference document containing all 19 skills with descriptions and example prompts. You can paste it directly into any AI CLI tool — including Ollama — to give the AI context about your educational workflow needs.

ollama run llama3.2 "$(cat docs/skills.md) — generate a lesson plan for Year 4 Science"

Or use it with Claude Code CLI:

/skill teaching-design-lesson-plan-generator
âŦ‡ Download skills.md View on GitHub →

đŸŽ¯ Practical Workflow: Rural Teacher Plans a Lesson Offline

Here is a step-by-step example of how a teacher in a rural Sarawak school can use AI to prepare a lesson without any internet connection:

  1. Open your terminal On Windows: press Win + R, type cmd, press Enter. On Mac: open Terminal from Applications.
  2. Start Ollama with a lightweight model Type: ollama run qwen2.5:3b — this works completely offline after the first-time download.
  3. Paste a skill prompt from skills.md Open docs/skills.md in Notepad, copy the "Lesson Plan Generator" example prompt, and paste it into the Ollama chat.
  4. Get your lesson plan output The AI generates a complete, structured lesson plan in seconds — no internet needed, no data costs.
  5. (When online) Use Claude Code CLI for richer results When back in a connected area, run /skill teaching-design-lesson-plan-generator in Claude Code for more sophisticated, curriculum-aware outputs.

💡 Pro tip: Download all four recommended Ollama models during school term breaks when you have reliable internet. Then you have multiple offline AI options ready for the entire semester.