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

Kenya has hundreds of smart agriculture platforms. Most of them break when it rains.
Unstable power, dropping 4G, and thousands of sensors reconnecting simultaneously after an outage are not edge cases. They are the baseline.
This session tells the story of an architecture that broke under those conditions and what it took to fix it — decision by decision, tradeoff by tradeoff. The result is a reusable pattern for any precision agriculture platform building on IoT in Kenya.
Ingestion — AWS IoT Core over MQTT protects constrained device battery life. Kinesis absorbs reconnection spikes without cascading compute failures.
The Edge Layer — ESP32-class devices buffer locally and publish on reconnection. A gateway running AWS IoT Greengrass filters data before it reaches the cloud.
Spatial Processing — Shapely inside Lambda handles spatial filtering and geofencing without dedicated GIS infrastructure.
Hot and Cold Paths — Kinesis fans out to Lambda, writing live state to RDS PostgreSQL with PostGIS, and Firehose batching records to S3 for Athena analytics.
Data Residency — AWS IoT Core is not available in af-south-1. This talk addresses what that means under the Kenya Data Protection Act and how the architecture responds.
Live Simulation — LocalStack and k6 show how synchronous ingestion collapses under a reconnection flood and how the event-driven pipeline absorbs the same load.