muend.
Open source · active
GeoAI · Spatial ML · Agent Systems Portfolio / 2026

Muhammed
Enes Duran I turn geospatial methods into reliable software systems.

I build production-grade GeoAI: agent skill systems, secure ArcGIS automation, reproducible Sentinel-2 pipelines, spatial decision-support products, and applied simulations designed around the failure modes that ordinary software quietly ignores.

Explore GeoAI Skills Open live DSS GitHub
Agentic GIS / MCP infrastructure Remote sensing / Sentinel-2 Spatial ML / leakage-aware validation Decision systems / browser-native delivery Applied simulation / rules into interaction
18
GeoAI agent skills
100
ArcGIS tools
120
Frozen routing cases
92.5%
Full-route accuracy
147
Neighborhoods modeled
2209-A
TÜBİTAK funded
01

Flagship Systems

05 systems

Not isolated demos. Each project occupies a specific layer of a broader spatial software stack: knowledge, execution, data preparation, decision delivery, or simulation.

Agent Skills · GeoAI knowledge infrastructure
MIT18 skillsactive

geoai-skills

A complete geospatial discipline layer for AI agents: acquisition, remote sensing, spatial statistics, deep learning, geostatistics, network analysis, cartography, PostGIS, and guarded ArcGIS execution.

  • Fail-loud defaults for CRS, units, geometry validity, leakage, and uncertainty.
  • Orchestrated multi-skill routing across the full GeoAI lifecycle.
  • Versioned evaluation harness with deterministic routing evidence.
18specialist skills
100%routing precision*
92.5%route accuracy*
Repository ↗Benchmark ↗
01 / flagship
MCP · Secure GIS execution
PyPI100 tools

arcgis-mcp-bridge

A local-first MCP framework that exposes ArcGIS Pro geoprocessing while isolating the licensed ArcPy runtime behind a validated worker boundary.

  • Async MCP server with a per-job ArcPy subprocess worker.
  • PathGuard enforced independently in both processes.
  • Structured failures and an 86-test offline quality gate.
LOCAL-FIRST MCP / TWO-PROCESS ARCHITECTURE SHEET B-02 · VERIFIED AGAINST REPOSITORY 01 / MCP HOST CLAUDE DESKTOP · CURSOR · OTHER MCP HOSTS JSON-RPC 2.0 over stdio — untrusted tool arguments enter here STDIO LAYER A / BRIDGE INTERPRETER · arcgis_mcp/server.py FASTMCP SERVER ASYNC DISPATCH · NO ARCPY IMPORT CONTRACT + REGISTRY PYDANTIC V2 · 100 TOOLSPEC SAFETY FLOOR A PATHGUARD PRE-CHECK · CONFIRM GATE NDJSON SUBPROCESS BRIDGE · SPAWNED PER JOB LAYER B / LICENSED ARCGIS PRO INTERPRETER · arcgis_mcp/worker.py PATHGUARD B RE-VALIDATE RUN_TOOL / ARCPY WORKER ARCPY IMPORT LEGAL ONLY HERE ARCGIS PRO RUNTIME .APRX · .GDB · RASTER · NETWORK FINAL NDJSON RESULT FRAME · VALIDATION / SECURITY / LICENSE / GEOPROCESSING / INTERNAL
Repository ↗PyPI ↗
02
Remote sensing · Reproducible data pipeline
PyPIZenodo DOI

sentinel-crop-pipeline

Sentinel-2 L2A preparation for crop classification—from CDSE discovery and AOI-cropped download to cloud-aware preprocessing, spectral indices, spatially blocked patches, and aligned ground-truth masks.

  • Cloud, shadow, snow, and invalid-pixel discipline.
  • COG/TIFF plus NPY/TFRecord training exports.
  • Spatial blocking designed to reduce train/test leakage.
Repository ↗PyPI ↗DOI ↗
03
Spatial DSS · Public-facing product
live147 areas

agri-dss

A zero-backend decision-support system for the Western Antalya agricultural corridor, translating crop and regional knowledge into practical neighborhood-level plans.

  • Five districts and 147 neighborhoods.
  • Auditable data-driven recommendation logic.
  • Clean A4 field plans and static deployment.
WESTERN ANTALYA / AUDITABLE DECISION CHAIN 5 DISTRICTS · 147 NEIGHBORHOODS 01 / TERRITORIAL INDEX DEMRE · FİNİKE · KAŞ · KEMER · KUMLUCA Neighborhood selection resolves a compact, versioned region and crop data contract. 02 / RECOMMENDATION LOGIC CLIMATE WATER SOIL SLOPE MARKET SCHEMA-DRIVEN SCORE · TRACEABLE INPUTS · NO SERVER-SIDE BLACK BOX 03 / NEIGHBORHOOD PLAN 01 · OLIVEHIGH FIT 02 · CITRUSFIELD CHECK A4 ACTION PLANPRINT / SHARE
Live app ↗Repository ↗
04
Applied simulation · Browser game
published

FOUNDER.EXE

A startup simulation that encodes taxes, incorporation, grants, investors, regulation, runway pressure, bankruptcy, and exit logic for Türkiye and the USA.

  • Idea-aware sector and market assumptions.
  • Optional AI advisor through player-owned API keys.
  • Static, browser-native release with local saves.
FOUNDER.EXE / RULE-DRIVEN OPERATING MODEL TÜRKİYE · USA · LOCAL SAVE 01 / RULESET COUNTRY + COMPANY TYPE TAX · GRANTS · REGULATION MONTH 08 IDEA-AWARE SECTOR PATH SCRIPTED OFFLINE FALLBACK ADVANCE MONTH 02 / OPERATING STATE RUNWAY / CASH 5.2 MONTHS BURN ↑ · REVENUE → · TAX DUE 18D CAP TABLE FOUNDER 58% · INVESTOR 28% · OPTION POOL 14% 03 / DECISION + CONSEQUENCE HIRE PIVOT COMPLIANCE DEADLINE REGISTER · PAY · ACCEPT RISK OUTCOME ENGINE SURVIVE · FAIL · FUND · EXIT
Play on itch.io ↗
05

* Published routing metrics apply to the frozen 17-skill, 120-case Claude Code / Claude Sonnet suite; they are not claims about general answer quality.

02

Evidence, Not Claims

auditable outputs

The portfolio is built around artifacts that can be inspected: packages, deterministic evaluations, validation scripts, DOI-backed releases, live products, and explicit limitations.

Routing benchmark

Strict full-route accuracy on the published frozen suite.

v1 / evidence
92.5%full route
120frozen cases
17benchmarked skills
Published result card

Measured routing behavior

Deterministic scoring from recorded activations, with an enabled/disabled control and explicit scope limitations.

Open evidence
Precision
100%
Recall
92.86%
Route accuracy
92.5%
PyPI × 2installable research / infrastructure packages
Zenodo DOIcitable Sentinel-2 pipeline artifact
Live systemsDSS and simulation shipped in browser
01 / Spatial rigor

CRS and units are explicit

Area, distance, raster alignment, scale, and geometry assumptions are checked before methods are trusted.

02 / Honest validation

Leakage is treated as a risk

Spatially dependent data requires blocked splits, appropriate baselines, and validation designs that reflect deployment.

03 / Reproducibility

Evidence is versioned

Typed contracts, CI, immutable benchmark packages, test harnesses, and DOI-backed releases keep work reviewable.

04 / Product utility

Methods become usable systems

Analysis is translated into interfaces, reports, local automation, or playable simulations rather than left as notebooks.

03

Academic & Research

03 studies

Research work focused on remote-sensing segmentation, spatial econometrics, and composite territorial indicators.

R / 01

agri-unet

Deep learning · Earth observation

Downstream agricultural pattern identification from satellite imagery, separated from the reusable data-preparation layer so experiments remain auditable.

 R / 02

turkiye-housing-prices-pandemic

Spatial econometrics

Inflation-adjusted housing-price analysis with regional comparisons and local spatial statistics for significant clusters and outliers.

 R / 03

kutri-resilience-index

Composite indicators · Urban resilience

A reproducible five-pillar territorial resilience index prototype with transparent normalization, weighting, figures, and methodology.
04

How I Build

four-stage discipline

The same sequence governs infrastructure, research software, decision systems, and simulations.

Phase 01

Define invariants

Make CRS, units, data contracts, safety boundaries, and success criteria explicit before implementation.

Phase 02

Build the pipeline

Separate components, isolate risky runtimes, preserve provenance, and make intermediate states inspectable.

Phase 03

Stress the failure modes

Test negative cases, leakage, invalid paths, ambiguous routing, missing data, and operational boundaries.

Phase 04

Ship a usable artifact

Deliver a package, live interface, citable release, benchmark, print output, or deployment-ready system.

05

Technical Core

capability matrix

A spatial-first stack spanning research methods, agent infrastructure, production engineering, and browser-native product work.

Spatial / remote sensing

Geospatial methods with operational discipline.

Vector, raster, EO, spatial statistics, cartography, and proprietary GIS workflows handled with explicit projections, scale, validation, and output checks.

Primary domainGeoAI
Core failure modeSilent spatial error
DeliveryPackage · map · DSS
GeoPandasShapelyArcPyRasterioPySALQGISSentinel-2CDSESTACPostGIS
06

System Map

one connected practice

The work compounds: methods inform infrastructure, infrastructure enables execution, and both feed products that can be used outside a research notebook.

Knowledge layer

geoai-skills

Routes spatial work and encodes methodological invariants.

Execution layer

arcgis-mcp-bridge

Executes licensed ArcGIS operations behind guarded boundaries.

Core practice

Reliable spatial systems

Research rigor, engineering discipline, and product delivery in one workflow.

Data / model layer

Sentinel pipeline + U-Net

Prepares reproducible EO datasets and downstream segmentation work.

Product layer

agri-dss + FOUNDER.EXE

Turns spatial or institutional rules into usable interactive systems.

Open to collaboration

Build spatial systems
that hold up.

Available for collaboration around GeoAI agent systems, production-grade spatial data science, remote-sensing ML pipelines, GIS automation, decision-support products, and applied simulations.

edcoders@gmail.com github.com/muend PyPI profile

Based in Türkiye · Open-source, research, and production collaboration welcome.