At Vectron Systems, I lead end-to-end delivery of AI-driven hardware products for autonomous sensing. I architected an edge-to-cloud inference system for 200+ sensor nodes, keeping pipeline latency below 120 ms at the 95th percentile under peak load.
I also built a React and FastAPI monitoring application that lets field engineers update models and review sensor health remotely. A Prometheus alert caught sensor contamination six hours before the customer's QA process would have flagged it.
Previously, at Lumora AI GmbH, I ported a defect-detection model to ONNX and TensorRT for NVIDIA Jetson, reducing inference latency from 340 ms to 74 ms. At Fieldnode Technologies, I developed IoT data services and sensor-processing scripts that improved data availability and reduced false anomaly alerts.

