Edge AI & IoT for Africa

Course outline and learning expectations

This is a tutor-led course. The outline shows what your tutor will cover; teaching materials and examinations are provided directly to enrolled students.

What you will learn

• Explain edge AI and identify when local inference is more suitable than cloud-only processing
• Connect, calibrate and read IoT sensors for soil moisture, weather and water-level applications
• Prepare sensor data and design a basic real-time inference pipeline
• Deploy a compact AI model to a supported microcontroller or Raspberry Pi device
• Apply quantization and pruning, then compare model size, memory use, latency and accuracy
• Connect sensor readings to AI predictions, rules, alerts and local device actions
• Design a power-aware prototype using duty cycling, sleep modes and solar or battery power
• Implement local buffering, store-and-forward synchronisation and recovery after connectivity loss
• Build, test and document an edge AI and IoT capstone prototype for smart irrigation

Modules

1. 1

   Edge AI for low-connectivity regions

   Understand on-device intelligence and select suitable edge or cloud processing for an African use case

   Edge AI conceptsEdge versus cloud inferenceLatency, privacy, bandwidth and reliabilityDevice constraintsSelecting an appropriate architecture
2. 2

   IoT sensors relevant to African applications

   Collect reliable physical-world data for agriculture, weather and water-monitoring prototypes

   Soil-moisture sensorsTemperature and humidity sensorsRainfall and weather-station inputsWater-level sensingSensor calibrationNoise, missing readings and data quality
3. 3

   Running AI models on microcontrollers and Raspberry Pi

   Prepare and deploy compact models to embedded and Linux-based edge devices

   Microcontroller versus single-board computerLiteRT for MicrocontrollersLiteRT on Raspberry PiExecuTorch overviewModel conversion and deploymentOn-device inference testing
4. 4

   Model compression for edge devices

   Reduce model resource requirements while measuring the effect on performance

   Post-training quantizationQuantization-aware training overviewWeight pruningModel size and memory measurementInference latencyAccuracy trade-offs
5. 5

   Connecting sensors to AI models for real-time decisions

   Build a sensor-to-decision pipeline that produces local alerts or device actions

   Sensor samplingFeature preparationInference loopsDecision thresholdsLocal alerts and actuatorsBasic device authentication and safe data handling
6. 6

   Power-efficient solar and battery design

   Plan an energy-aware prototype suitable for remote or off-grid operation

   Power budgetingDuty cyclingMicrocontroller sleep modesEfficient sensor samplingSolar charging basicsBattery protectionEnergy-use testing
7. 7

   Offline-first data synchronisation

   Keep the device useful during connectivity loss and synchronise data when a connection returns

   Local storageTimestampingData queuesStore-and-forward patternsRetry and back-off logicDuplicate handlingData integrity after reconnection
8. 8

   Capstone - solar-powered soil-moisture and irrigation-alert device

   Design, assemble, test and document an edge AI and IoT prototype for a smallholder farm

   Problem definitionHardware selectionSensor calibrationModel or decision-logic developmentOn-device deploymentLow-power testingOffline data loggingIrrigation alertsPrototype evaluation and demonstration