Calming the Storm: Building Resilient IoT Infrastructure for Precision Agriculture in Kenya

Precision agriculture platforms in Kenya must operate through unstable power, intermittent 4G connectivity, and thousands of sensors reconnecting simultaneously after an outage. These are not edge cases; they are normal operating conditions.

This session follows an architecture that failed under those conditions and the decisions, tradeoffs, and redesign that made it resilient. The result is a reusable event-driven pattern for IoT platforms operating in constrained environments.

Architecture

• Ingestion: AWS IoT Core uses MQTT to protect constrained-device battery life, while Kinesis absorbs reconnection spikes without cascading compute failures.
• Edge layer: ESP32-class devices buffer readings locally and publish after reconnecting. AWS IoT Greengrass gateways filter data before it reaches the cloud.
• Spatial processing: Shapely in AWS Lambda provides spatial filtering and geofencing without dedicated GIS infrastructure.
• Hot path: Lambda writes live state from Kinesis to RDS PostgreSQL with PostGIS.
• Cold path: Firehose batches records into S3 for analysis with Athena.
• Data residency: The design addresses the absence of AWS IoT Core in af-south-1 and its implications under Kenya’s Data Protection Act.
• Live simulation: LocalStack and k6 demonstrate how synchronous ingestion collapses during a reconnection flood and how the event-driven pipeline absorbs the same load.

Format: 30-minute technical session
Level: Intermediate to advanced
Demo repository: ZEZE1020/aws-cdk-demo