Rafi T.

Wrong parcel lines and outdated map data cost real money. I'm Rafi, a WebGIS developer and GIS data specialist (M.Sc. TU Berlin, ex DLR) who extracts, verifies and maps location data with high accuracy, especially for real estate and parcel research. I turn messy sources like PDFs, public records and web data into clean datasets, accurate maps, interactive WebGIS dashboards and automated workflows, with machine learning and AI automation when the job needs it. All five of my client reviews are 5 stars, and one repeat client hired me twice for map data accuracy work. Who I help • Real estate and parcel researchers • Startups and product teams building map based apps, dashboards and AI features • Data and analytics teams • Urban planners • Telecom and infrastructure companies • Environmental and forestry teams • Agriculture and agritech teams • Insurance and risk analysts • NGOs and consultancies • Research groups Background M.Sc. in Geodesy and Geoinformation Science (TU Berlin) with GIS, remote sensing and machine learning experience, including the German Aerospace Center (DLR) and OXG Glasfaser GmbH. Recent results • Riverfront parcel research for 1,139 properties, delivered as Excel/CSV plus an interactive map • Extracted 100,000+ tree crowns from stereo aerial imagery with deep learning (YOLOv12l, F1 0.73) at DLR • Automated geospatial analysis across 20+ cities for fibre network rollout, improving efficiency by 20% • Analyzed 10 years of satellite imagery that revealed 15% forest loss Upwork and freelance GIS projects • Ongoing map data accuracy work for a repeat client, hired twice • Riverfront parcel research for 1,139 properties, delivered as Excel/CSV plus an interactive map • GIS data extraction, GeoJSON mapping from PDF sources, and heatmap design for 10x10 km grid data • Forest change detection map app with a Streamlit dashboard • Sentinel 1 SAR and GIS Random Forest flash flood susceptibility map for Taif (ROC AUC 0.82), with a fully reproducible pipeline and a conference paper • PDF summary chatbot with retrieval augmented generation (RAG) for a plantation project What I deliver • GIS data extraction and web scraping with Python, delivered as clean Excel and CSV tables (requests, BeautifulSoup, lxml, pandas) • Map data accuracy review and quality checks, verifying boundaries, attributes and sources against official records • WebGIS and interactive maps, dashboards and map APIs with a spatial database backend (Leaflet, Mapbox, Folium, Streamlit, FastAPI, PostGIS, GeoJSON) • GIS analysis, map digitizing, georeferencing, geocoding, and cartography (QGIS, ArcGIS Pro) • Spatial and proximity analysis (buffers, isochrones, network routing, overlays, spatial joins) • Geospatial data pipelines, spatial databases, ETL, and workflow automation (PostGIS, Python, Docker) • Data science and machine learning, including data cleaning, exploratory analysis, regression, classification, clustering and predictive models (Python, pandas, scikit-learn, XGBoost) • Computer vision and deep learning for imagery, including image annotation, dataset labeling, and object detection of tree crowns and crop diseases (PyTorch, YOLO) • Hands on model development, including manual model training, hyperparameter tuning, testing, cross validation, accuracy assessment (precision, recall, F1, mAP, ROC AUC), error analysis and performance monitoring • AI automation, including document chatbots with retrieval augmented generation (RAG), natural language to spatial SQL tools, and automated data workflows (Python, FastAPI, Docker) • Spatial statistics and cluster analysis • Remote sensing and satellite image analysis, change detection, and land cover classification (Google Earth Engine, Rasterio, GDAL, Sentinel 1 and 2, Landsat) • LiDAR point cloud processing, DEM, DSM and canopy height models, and tree extraction (PDAL, laspy, Python) You receive Trained models with documented accuracy metrics, clean code, GeoJSON, shapefiles, GeoPackages, GeoTIFFs, Excel and CSV tables, KML/KMZ, print ready PDF maps, and interactive web maps or dashboards with a live link, with the workflow, assumptions, and quality checks documented so results are reproducible. To get started, send me your source data, the outputs you need, and your deadline. I reply within 6 hours and will recommend a workflow before we begin.