PilotEdgePredictive
TerraTech Agri Insight
ESP32 nodes with FastAPI backend, telemetry, and Random Forest predictions.
IoT Engineer
IoT & R&DESP32MQTTFastAPIInfluxDBRandom Forest
Context
Problem & context
- Farms needed resilient telemetry with buffering during outages.
- Wanted predictive insights without heavy cloud cost.
Architecture
Architecture & stack
- ESP32 sensor nodes with store-and-forward MQTT buffering.
- FastAPI backend ingesting telemetry into InfluxDB.
- Random Forest model scoring moisture and environment data.
Features
Key flows
- Telemetry dashboards with historical slices.
- Predictive irrigation suggestions.
- Alerting on sensor drift and offline nodes.
Challenges
Engineering challenges & solutions
- Reliable buffering when connectivity drops.
- Model drift when seasons change.
- Keeping power usage low on edge devices.
Impact
Impact & metrics
- Increased data reliability with store-and-forward design.
- Early wins in water optimization from predictive model.
- Template usable for other agri nodes.
Next
What I’d improve next
- Auto model retraining per season.
- Edge inferencing to cut latency further.