I build production-grade GeoAI, spatial decision-support systems, and applied simulation products: ArcGIS automation exposed to LLM agents, deep-learning pipelines for satellite imagery, browser-native DSS tools, and realistic strategy simulations that turn complex real-world rules into usable software.

I focus on structural data science, spatial machine learning, and applied product systems. My work connects rigorous mathematical models with interfaces people can actually use: from GIS automation layers and satellite-imagery pipelines to static, browser-native decision-support and simulation products.

Core AI / ML

Spatial Data Science

Systems, MLOps & Product Engineering

An institutional-grade Model Context Protocol (MCP) framework exposing exactly 100 specialized geoprocessing tools directly to LLM hosts and intelligent agents — turning ArcGIS Pro into a programmable backend for AI workflows.

• Process Isolation: Strictly decoupled multi-process architecture: async MCP server core and isolated ArcPy worker subprocess.
• Security Layer: PathGuard sandbox with prefix validation over filesystem and geodatabase paths before algorithmic execution.
• Distribution & Verification: Published on PyPI, listed on Glama with A-rated MCP quality, and validated through 81 passing tests with mocked ArcPy.
• Agentic GIS: Designed for local, controlled ArcGIS Pro automation from LLM hosts without collapsing the LLM runtime into the licensed GIS interpreter.

flowchart LR
    A["LLM Host / AI Agent"] -->|MCP protocol| B["MCP Server<br/>Async Core"]
    B --> C{"PathGuard<br/>sandbox"}
    C -->|path validated| D["Isolated Worker<br/>Subprocess"]
    C -.->|rejected| X["Blocked"]
    D --> E["ArcGIS Pro · ArcPy"]
    E --> F["100 geoprocessing tools"]

A fully client-side Spatial Decision Support System for the Western Antalya agricultural corridor — 5 districts, 147 neighborhoods (Demre, Finike, Kaş, Kemer, Kumluca) — turning local agronomic and economic knowledge into a concrete, printable plan for each neighborhood.

• Zero-Backend Static Architecture: Vanilla JS application against a decoupled data.json layer — no backend, no build step, trivially hostable and auditable.
• DRY Data Contract: Compact cropSets / longTermCrops / regions schema resolved at runtime; recommendations can be updated by editing data only.
• Swiss / Typographic Interface: Guided stepper, corridor diagram, live counters, and clean A4 print output for village boards and cooperatives.

A browser-based startup simulation game modeling the practical stress of founding a company in Türkiye or the USA: taxes, incorporation choices, grants, investor expectations, cash flow, regulatory friction, AI-assisted advising, and bankruptcy/exit outcomes.

• Rules-Driven Simulation: Country-specific rule sets for Türkiye and the USA, including tax filings, company types, grants, regulatory constraints, inflation/currency pressure, investor interactions, and runway management.
• Idea-Aware Onboarding: The player writes a startup idea; the system analyzes sector, regulatory requirements, global potential, and market path before shaping the simulation.
• Optional Live AI Layer: Gemini, OpenAI, or Anthropic can be connected with the player's own API key; the game remains playable with scripted fallbacks when offline.
• Browser-Native Product: Built as a static web simulation with local save persistence, rule-based economic modeling, country-specific startup/legal parameters, and optional LLM-powered advisory flows.

Reproducible, peer-review–oriented studies in spatial econometrics, urban resilience, and deep learning for remote sensing.

The codebase for my TÜBİTAK 2209-A research project (University Students Research Projects Support Program) — an accepted, grant-funded study. A U-Net semantic segmentation pipeline for agricultural pattern identification from high-resolution, multi-temporal satellite imagery, extracting field parcels and crop structures for downstream suitability modeling.

Region-level analysis of Türkiye's housing market that separates real (inflation-adjusted) price growth from inflation, comparing the six years before and after the COVID-19 pandemic (2014–2025).

• Reproducible Python notebook with high-resolution choropleth figures and a House Price Index (HPI) deflation pipeline.
• LISA (Local Indicators of Spatial Association) analysis to detect statistically significant regional clusters and spatial outliers.

A reproducible urban-territorial resilience index prototype for Kaş / Bayındır, Antalya, based on a five-pillar composite indicator framework.

• Transparent indicator normalization and weighting methodology with fully reproducible notebooks and figures.
• Bridges quantitative spatial analysis with applied territorial planning.

Start here depending on what you want to inspect:

These are the product narratives I am currently expanding into separate case studies:

• FOUNDER.EXE: encoding company formation, tax pressure, grants, investment logic, regulatory constraints, and optional AI advising into a playable browser simulation.
• Agri-DSS: compressing agricultural suitability and local economic knowledge into printable, village-level decision plans.
• arcgis-mcp-bridge: separating licensed GIS execution from AI host runtimes while preserving secure, discoverable, and quality-scored geoprocessing access.

• Computer Vision for Remote Sensing: Automated pipelines for agricultural pattern identification and urban object extraction from high-resolution multi-temporal satellite imagery using CNN and U-Net segmentation models.
• Agentic GIS & MCP Tooling: Secure local automation layers that expose GIS operations to LLM agents without compromising ArcGIS Pro runtime isolation.
• Applied Simulation Systems: Browser-native simulations that encode real-world regulatory, financial, and decision processes into interactive products.
• Urban Resilience Forecasting: Predictive spatial suitability matrices and long-term territorial resilience frameworks using robust statistical models.

Open to collaborative tracks involving production-grade Data Science, Spatial Machine Learning pipelines, automated GeoAI systems architecture, and applied simulation products.

• Portfolio: muend.github.io
• Live App: tarimsalkoridor.online
• Game: FOUNDER.EXE on itch.io
• Email: edcoders@gmail.com