Define invariants
Make CRS, units, data contracts, safety boundaries and success criteria explicit before implementation.
How I build
Practice
The same sequence governs infrastructure, research software, decision systems and simulations. It is ordered so that the expensive mistakes are made cheap.
StagesFour, always in order
Primary domainGeoAI and spatial data science
Failure modeSilent spatial error
Build the pipeline
Separate components, isolate risky runtimes, preserve provenance, and make intermediate states inspectable.
Stress the failure modes
Test negative cases, leakage, invalid paths, ambiguous routing, missing data and operational boundaries.
Ship a usable artifact
Deliver a package, live interface, citable release, benchmark, print output or deployment-ready system.
Principles
What that sequence is protecting against
CRS and units are explicit
Area, distance, raster alignment, scale and geometry assumptions are checked before methods are trusted.
Leakage is treated as a risk
Spatially dependent data requires blocked splits, appropriate baselines and validation designs that reflect deployment.
Evidence is versioned
Typed contracts, CI, immutable benchmark packages, test harnesses and DOI-backed releases keep work reviewable.
Methods become usable systems
Analysis is translated into interfaces, reports, local automation or playable simulations rather than left as notebooks.
Stack
Technical core
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
Working set10
GeoPandasShapelyArcPyRasterioPySALQGISSentinel-2CDSESTACPostGIS
Agents / evaluation
Tool-using agents that route, execute and fail safely.
Agent Skills, MCP, runtime isolation, typed evaluation contracts, blind request preparation, deterministic scoring and evidence-aware benchmark publication.
Primary domainAgentic GIS
Core failure modeUnsafe or wrong routing
DeliveryPlugin · MCP · benchmark
Working set08
Agent SkillsMCPJSON-RPCJSON SchemaRouting evalsBehavior evalsGuarded executionBenchmark harnesses
AI / ML / data
Models built on defensible data and evaluation.
Deep-learning and classical ML workflows with spatial splitting, class-imbalance handling, reproducible preprocessing and georeferencing-preserving inference.
Primary domainSpatial ML
Core failure modeLeakage / false confidence
DeliveryDataset · model · report
Working set07
PythonPyTorchTensorFlowscikit-learnNumPypandasSciPy
Systems / product
Research logic translated into software people can operate.
Typed Python services, static web products, CI, testing, distribution, browser-native interfaces, local persistence and low-maintenance deployment architectures.
Primary domainProduct engineering
Core failure modeUnmaintainable delivery
DeliveryPyPI · web · game
Working set10
FastAPIPydantic v2DockerPytestRuffMypyGitHub ActionsVanilla JSHTML/CSSWeb storage
Fig. 01
How the projects feed each other
The work compounds: methods inform infrastructure, infrastructure enables execution, and both feed products that can be used outside a research notebook.
© 2026 Muhammed Enes Duran
Static site · no trackers