Evaluate storm damage and dispatch maintenance crews

​After a storm passes, the crews that restore power need to know three things quickly: which poles are down, what broke on each one, and which crew to send first. Getting those answers might mean flying a drone, bringing the footage back, and having someone review it frame by frame before a single work order goes out.

In this session, Roboflow Engineers Jennifer Kuchta and Alexei Alexandrovich will demonstrate a storm response system built to shorten the time it takes to assess damage and create maintenance tickets. The system connects to a drone feed via Real-Time Messaging Protocol (RTMP), confirms the types of damage, and plugs the data into Salesforce as a maintenance request.

In this live session, we will cover:

• ​Building the model with limited data: How GPT-6 Astra and SAM 3 can label utility poles, transformers, and storm-torn vegetation automatically, turning a handful of clips into a trained model quickly and efficiently.

• ​Assessing damage, not just detecting it: The logic that reads a scene and works out what the repair will take, from a snapped pole to a downed transformer to a stump that still needs a ground crew.

• ​Dispatching the work automatically: Connecting the Workflow to Salesforce so each finding becomes a routed maintenance request, sequenced so crews arrive in the order the repair needs them.