📊 Projects & Technical Overview
Summary comparison of my core engineering and geospatial software projects
| Image | Project Name | Role & Domain | Core Technologies | Key Achievement |
|---|---|---|---|---|
 |
AI-Powered Georeferenced Road Extraction | Lead Developer • GeoAI | PyTorch (SMP U-Net), FastAPI, React, TypeScript, Rasterio (GDAL), OpenCV | 84.3% IoU pixel-level road extraction from 1GB+ GeoTIFF satellite imagery with interactive Dijkstra/A* repair |
 |
REZIT – SaaS Reservation Platform | Founder & Lead Dev • SaaS | Node.js, Express.js, PostgreSQL, React.js, SVG/HTML5 Canvas, JWT | 500+ req/sec concurrency, ACID pessimistic row-locking & interactive affine seat mapping (rezit.app) |
 |
Gordion Ancient City 3D Digital Twin | Field Survey Assistant • 3D GIS | GNSS / GPS Field Surveys, UAV Data Acquisition, Indoor LiDAR Scanning | Active field participation in GPS measurements, drone flights & museum indoor LiDAR scanning (dijitalikiz.hkmo.org.tr) |
 |
İzmir Disaster Gathering Areas Web GIS | Researcher • Web GIS | Web GIS, Spatial Network & Accessibility Analysis, QGIS, ArcGIS | Presented internationally at the 35th IGSM in Valencia, Spain |
 |
MERYKSI CO. – AI Print-on-Demand | Founder • E-Commerce | Generative AI Design, Etsy SEO Analytics, Printify Automation | 94+ global sales across 70+ cities and 3 continents |
🚀 Featured Engineering Projects
Lead Developer • Geospatial AI & Computer Vision
AI-Powered Georeferenced Road Extraction & Interactive Editing Platform
An end-to-end geospatial artificial intelligence platform that automatically extracts pixel-level road networks from high-resolution satellite imagery and enables real-time interactive topological repair.
- End-to-End GeoAI Segmentation: Trained a multi-encoder (EfficientNet-B3 / ResNet50) U-Net segmentation architecture using Segmentation Models PyTorch (SMP) with hybrid Dice + Focal loss and Albumentations, achieving an 84.3% IoU score.
- Interactive Road Repair & Tiling: Engineered an interactive repair module combining Test-Time Augmentation (TTA), Laplacian gradients, and Dijkstra / A* shortest-path algorithms to bridge fragmented road segments in real time. Implemented 512px overlapping sliding-window inference (stride: 128px) with probability stitching to process 1GB+ GeoTIFFs without GPU OOM errors.
- Lossless Georeferencing: Preserved CRS projections and Affine transformation matrices via Rasterio (GDAL) alongside an asynchronous FastAPI background task queue for seamless QGIS & ArcGIS compatibility.
Founder & Lead Developer (2025 – Present)
REZIT – Next-Gen SaaS Interactive Table Reservation Platform
A high-concurrency SaaS reservation platform that allows customers to interactively inspect venue layouts and select their exact table before arriving at the venue.
- Scalable Backend Architecture: Built from scratch with Node.js, Express.js, and PostgreSQL utilizing B-Tree indexing and
pg-poolconnection pooling to reliably handle 500+ concurrent requests/sec under peak load. - Transaction Safety & Interactive Floorplan Editor: Eliminated double-bookings and race conditions via ACID transactions and pessimistic row-locking (
SELECT FOR UPDATE). Designed an interactive React template editor and SVG/HTML5 Canvas seat-mapping module powered by affine matrix transformations (smooth zoom/pan). - Enterprise Security Infrastructure: Integrated stateless JWT authentication, strict CORS policies, Express Rate-Limiter layers, and Express-Validator endpoint sanitization against SQL injection.
Field Survey Assistant • HKMO Ankara (2024)
Gordion Ancient City 3D Digital Twin Project
Actively participated in on-site geodetic field operations for the 3D digital twin documentation of the UNESCO World Heritage site Gordion Ancient City and Gordion Museum.
- Field GPS / GNSS Measurements: Actively took part in field work and assisted the surveying team with high-precision GPS / Ground Control Point (GCP) measurements across the archaeological site.
- UAV (Drone) Flight Operations: Participated on-site during drone aerial imagery acquisition campaigns, observing and learning photogrammetric flight planning and field data collection workflows firsthand.
- Indoor Terrestrial LiDAR Scanning: Present during the indoor LiDAR laser scanning of the Gordion Museum interior, gaining hands-on field familiarity with 3D point cloud acquisition and digital twin methodologies.