GIS agents and copilots
Multi-step GIS tasks, tool use, spatial analysis, and instruction following.
Brooks Geospatial
Geospatial AI Evaluation & GIS Quality Assurance
Testing whether AI can perform real GIS work — accurately, reproducibly, and with defensible outputs.
I build realistic ArcGIS and QGIS test cases, expert reference outputs, objective scoring rubrics, and failure reports for GIS agents, mapping tools, geocoders, spatial-data pipelines, and remote-sensing systems.
3+ years professional GIS experience
ArcGIS / ArcMap
QGIS portfolio development
Python and spatial-data QA
1,500+ record database experience
FAA Part 107
What I evaluate
Spatial systems need evidence, not just plausible output.
Evaluation is designed around real GIS work, clear inputs, defensible references, and criteria people can inspect.
Maps, geocoders, and location search
Location results that need coordinate, match-quality, and spatial-tolerance review.
Spatial-data pipelines
Schema, joins, record counts, geometry, lineage, and repeatability across transformations.
Remote-sensing and visual outputs
Map and image outputs that must be analytically useful, clearly communicated, and traceable.
Method
A six-step evaluation method
Each phase preserves the evidence needed to tell whether a system’s output is useful, correct, and repeatable.
- 01
Define the task
- 02
Assemble the data
- 03
Build the reference output
- 04
Run the system
- 05
Score the result
- 06
Document failures and regression tests
Evaluation dimensions
Measure the spatial details that matter.
- Analytical correctness
- CRS and units
- Geometry and topology
- Schema
- Record counts and joins
- Spatial tolerance
- Cartographic quality
- Provenance
- Instruction following
- Uncertainty and refusal
- Reproducibility
Featured projects
Evaluation work in development
Each project page identifies its current status and avoids client data or unsupported results.
Live ArcGIS Dashboard In progress/spatial-data-qa-toolkit
Spatial Data QA Toolkit
A reusable quality-assurance toolkit for checking the spatial and tabular conditions that make GIS outputs defensible.
- CRS, units, extent, and spatial-tolerance checks
- Geometry, topology, schema, null, duplicate, and orphan-record review
- Documented workflow checks that can be rerun as data changes
In progress/geoeval-benchmark
GeoEval Benchmark
A benchmark design for testing whether AI systems can execute realistic GIS tasks with reproducible, reviewable results.
- Realistic ArcGIS and QGIS task design
- Expert reference outputs and objective scoring rubrics
- Failure reports and regression cases for repeatable evaluation
In progress/texas-geocoder-evaluation
Texas Geocoder Evaluation
An in-progress evaluation design for reviewing geocoder and location-search outputs against transparent spatial criteria.
- Address and location-query task design
- Coordinate, match-quality, and spatial-tolerance review
- Documented failures for retesting after system changes
Professional background
GIS analysis grounded in real spatial work.
Davis “Mac” Brooks progressed from GIS Analyst Intern to GIS Analyst at Capitol Market Research, where he performed market-area mapping and real-estate feasibility research.
His work included maintaining a database of more than 1,500 office, retail, and multifamily locations, with experience across ArcGIS/ArcMap, QGIS portfolio development, Python, and spatial-data QA.
Read the professional backgroundContact
Need an evaluation plan for a geospatial system?
Discuss a focused evaluation pilot, a GIS QA workflow, or a reference-output review.