Computer Vision Engineer (On Site) at Grupo Premo

Built and productionized CV pipelines across 8+ production lines, combining object detection (YOLO) and industrial anomaly detection (PatchCore, EfficientAD) with classical vision to catch incomplete/defective parts in real time. Implemented and fine-tuned OCR to extract serial numbers and dates at scale, achieving ~95% accuracy and reducing manual checks by ~60%. Designed line-specific Python GUIs wrapping inference and QC logic for 3-click operator workflows, cutting inspection time by ~30%. Led end-to-end model lifecycle (data collection, labeling, training, evaluation) maintaining ≥80% accuracy and reducing false positives by ~25%. Integrated cameras via vendor SDKs and connected systems over TCP/IP, UDP, WebSockets, and RS232 to sustain real-time AOI inference at the edge. Built a Real-Time Quality Monitoring System aggregating AOI outputs and visualizing OK/Scrap; delivered mini-SCADA system for temperature/humidity monitoring with alarms and Teams webhook integration.