Rafi

GIS Analyst

DLR • Meio período

Jan 2026 - Present • 8 mos

• Developed a dual-backbone deep learning fusion pipeline in PyTorch Lightning for urban tree health monitoring using stereo aerial RGBI imagery. • Reduced class imbalance (26:1 → 15:1) through label cleaning and recovering under-detected declining trees, and applied weighted sampling and focal loss to improve minority-class F1 by 65% (0.20 → 0.33). • Built an automated YOLO detection pipeline with SAHI tiling to generate tree crown annotations from large GeoTIFF surveys, with YOLOv12l reaching an F1 of 0.73. • Constructed geospatial preprocessing pipelines for spatial joins, CRS-consistent raster cropping, and multi-channel patch extraction. • Benchmarked encoder and fusion configurations to find optimal setups. • Performed data quality checks and geospatial validation using QGIS and Python to ensure dataset integrity.

GIS Analyst

OXG Glasfaser GmbH • Meio período

Nov 2024 - Nov 2025 • 1 yr

• Improved operational efficiency by 20% through automated geospatial analysis using QGIS and Python to support fiber network deployment. • Performed cost-benefit analyses and site assessments to identify optimal deployment areas while navigating geographical constraints and market competition. • Developed Python-based geocoding tool for German addresses, reducing processing time by 30%.