ELDAD RON

Led end-to-end vision projects, employed comprehensive data strategies for data collection, EDA, and preprocessing. Developed efficient machine learning pipelines using GCP and Vertex AI. Collaborated with cross-functional teams to deliver advanced machine learning solutions.

• Granulometry Model: Harnessed classical computer vision algorithms to devise a tabular dataset, boosted bean size prediction accuracy by 20% using tree-based machine learning algorithms.
• Segmentation Model: Improved bean segmentation and reduced computation time and memory usage by 60%. Managed data collection, trained several algorithms, and implemented tiled inference.
• Defect Classification Model: Successfully managed project lifecycle, from conception to deployment, yielding a system that elevated quality of coffee beans.
• Deployment & Infrastructure: Directed models deployment through endpoint services and microservices. Designed a robust training infrastructure, facilitating smooth operations and efficient model utilization.

Implemented deep learning and computer vision algorithms. Collaborated with agronomists to enrich datasets and translated complex results into comprehensible insights, driving data-based decision making.

• Classification: Built advanced tiled classification models using Pytorch-Lightning, targeting plant diseases detection in high-resolution 4K images. Optimized and tuned Hyperparameters.
• Object Detection: YOLO-based system for flowers, insects, and plant disease detection.
• Instance Segmentation: Detectron2 based system for grape cluster instance segmentation.