Fabric Studio · Invite-Only Beta
Geospatial workflows weren't built to be portable. Studio is the visual workspace where you compose a Pattern — a portable representation of a pipeline, separated from the tools that execute it — and submit it to Fabric Engine. Build visually, run anywhere, walk away with every byte.
We onboard in small cohorts as capacity allows — typically 2–5 new users per week.
The problem
Every mission starts from zero. Analysts rebuild the same ingestion and harmonization plumbing across every project because the last pipeline lives in a notebook that only ran once, on one machine, with one analyst's assumptions baked in.
Studio inverts that. You compose the pipeline visually, and the workspace writes a Pattern — a declarative description of the whole workflow, portable across environments. The Pattern travels with you to the next mission, the next team, the next deployment. You keep your judgment; Fabric removes the plumbing.
The interface
A collapsible tree presenting all six Fabric layers (Area of Interest, Data Sources, Preprocessing, Analysis, Post-processing, Output) as folder groups. Each task shows its natural language name, the underlying function identifier in monospace, a toggle state (on / off / required), and an active count badge per layer. The left edge of each task highlights to indicate the current selection.
Click any task to load its configuration form in the center column. Breadcrumb navigation shows your position in the layer hierarchy. Inline code displays current values. Contextual info boxes (tips, success confirmations, and amber warnings) surface guardrail feedback as you configure. Previous / Next navigation moves through the layer sequence without losing state.
A vertically-oriented DAG that updates in real time as tasks are toggled on and off. Function names appear in monospace, the authoritative execution identifiers. Left border colors are keyed to the layer system. Parallel tasks render side-by-side. Clicking any node loads its configuration. A live Pattern preview at the bottom shows exactly what will be submitted to Engine.
The workflow
Start from a blank workspace and toggle tasks on layer by layer, or open one of the example Patterns for fire and wildfire, vegetation analysis, terrain, flood response, or general-purpose workflows. When Iris is available, you will also be able to describe your goal in plain language and get a generated starting Pattern.
Your starting Pattern loads into a visual workspace with tasks pre-configured. Every configuration is editable: adjust, extend, or strip back as needed. The Process Map on the right updates in real time as you make changes.
Toggle the Study Area Generator on and enter a location name — Studio geocodes it to a Bounding Box automatically. Or toggle Administrative Boundaries to pull a country, state, or county polygon from OpenStreetMap. The AOI propagates to every downstream task as the spatial scope.
Toggle data tasks on from the available sources: Sentinel-2, Sentinel-1 SAR, Landsat, SRTM, Copernicus DEM, OpenStreetMap vectors, ERA5, and HydroSHEDS. Set Date Ranges, cloud cover thresholds, and source collection overrides per source. The Process Map updates with each toggle.
Clip, Reproject & Resample, Align to Grid / Mosaic, and Normalize Nodata. Studio's guardrails watch your configuration: mixed-resolution inputs trigger an auto-recommended resampling task. Coarse-resolution sources alongside 10-meter Sentinel-2 auto-escalate grid alignment to required. Configuration conflicts surface as amber warnings before you run.
Choose from the spectral index library (NDVI, NDSI, NDWI, NBR, dNBR, NDBI, MNDWI), build Band Math expressions using named templates or write custom expressions, run NDVI (Harmonized Vegetation Analysis in Studio) across multi-sensor inputs, or run SAR Flood Detection. The band picker grays out unavailable indices and backfills band suggestions automatically.
Review the Pattern in the Process Map. This is exactly what Engine will execute. When ready, click Run. Studio submits to Engine and opens the live execution panel. Two progress bars show overall pipeline progress and per-task sub-steps in real time. Logs stream with step labels as each container completes.
Download your Pattern as YAML for version control, programmatic execution, or to hand off to a teammate. Analysis-ready outputs — Cloud-Optimized GeoTIFF (COG), GeoJSON, GeoPackage (GPKG), JSON reports, and a W3C PROV provenance manifest — are immediately compatible with ArcGIS, QGIS, Felt, Python, R, and any OGC-compatible tool.
What you keep
Scripts don't travel. Patterns do. Studio preserves the analytical judgment your scripts already encode — band choices, thresholds, composite logic — and replaces the boilerplate around them with a portable, provenance-tracked pipeline. Drop back to code whenever you want: export the Pattern as YAML, or run custom expressions inline.
Studio recommends; you decide. Every task is editable, every guardrail is advisory, every configuration is visible in the Process Map before a Pattern is submitted. W3C PROV provenance documents every transformation line-by-line, so you always know exactly what ran and why.
Outputs land in your storage as standard open formats — Cloud-Optimized GeoTIFF (COG), GeoJSON, GeoPackage (GPKG), W3C PROV manifests. No proprietary wrappers, no data tax. If you walk away, you walk away with every byte, every Pattern, every manifest.
Intelligent guardrails
Studio surfaces configuration guidance while you're composing the workflow, before anything is submitted to Engine. These are not post-hoc error messages — they appear inline as you configure.
A resolution pre-check blocks progression if band inputs have mismatched Coordinate Reference System (CRS), dimensions, or resolution. If mixed-resolution indices are selected (for example, NDMI requires both 10-meter and 20-meter Sentinel-2 bands), Studio auto-recommends a resampling task. If coarse-resolution sources are added alongside 10-meter data, grid alignment is automatically escalated to required status. Adding a new data source after an alignment task is already configured triggers a flag to extend it. The band picker grays out indices that aren't compatible with your current source selection and offers backfill suggestions. Named Band Math templates (NDVI, NDMI, NBR, dNBR) automatically resolve the correct upstream output paths without manual wiring.
The result: a pipeline that works the first time, without learning which preprocessing steps are prerequisites for which analysis tasks.
Spectral indices
Normalized Difference Vegetation Index.
(B08 − B04) / (B08 + B04)
Quantifies live green vegetation density. Inputs auto-wired
from upstream Sentinel-2 preprocessing output.
Normalized Difference Snow Index.
(B03 − B11) / (B03 + B11)
Discriminates snow and ice from cloud cover and other bright surfaces.
Water body and flood extent mapping.
NDWI: (B03 − B08) / (B03 + B08)
MNDWI: (B03 − B11) / (B03 + B11)
Burn severity and fire scar mapping.
NBR: (B08 − B12) / (B08 + B12)
dNBR: NBR_pre − NBR_post for change detection between two acquisition dates.
Normalized Difference Built-up Index.
(B11 − B08) / (B11 + B08)
Identifies built-up and impervious surfaces for urban growth and construction monitoring.
Generic templates for normalized difference, weighted composite, [0,1] normalization, conditional masking, and fully custom expressions. Manual input wiring for non-standard source combinations.
Where outputs go
Studio sits in the middle stack: ingestion, harmonization, preprocessing. The deep analysis and cartography happen downstream, in the tools your team already trusts. Fabric is the enabler that delivers analysis-ready data into those environments — not a replacement for them.
A Cloud-Optimized GeoTIFF from Studio loads into QGIS or ArcGIS as easily as any local raster. A GeoPackage opens directly in Felt for collaborative mapping. Python and R read the same outputs via rasterio or terra. The W3C PROV manifest travels with the data, so the provenance story is intact wherever the analysis lands.
FAQ
Studio is in Invite-Only Beta. Fabric Engine, the runtime underneath, is live and processing data today. We'll be honest about whether your use case is one we can support well right now — rather than overpromising and backfilling later.
Cloud-Optimized GeoTIFF (COG) for raster outputs, GeoJSON and GeoPackage (GPKG) for vector outputs, JSON for statistical reports, and a W3C PROV manifest per run. Everything loads directly into ArcGIS, QGIS, Felt, Python, R, or any OGC-compatible tool — no format conversion required.
No. Studio is no-code by default: toggle tasks, fill in forms, run. When you need code — custom Band Math, a bespoke transformation — drop into expression fields or export the Pattern as YAML and run it programmatically. Your choice, per workflow.
Geospatial analysts and engineers who have rebuilt an ingestion pipeline more than once and know exactly which hours they want back. We onboard in small cohorts — typically 2–5 new users per week as capacity allows — to keep the feedback loop tight.
Under the hood
Every workflow built in Studio generates a Pattern and submits it to Engine for execution. The containerized isolation, analysis-ready data guarantees, and manifest.json provenance record are Engine features. Studio exposes them through a visual interface and exports your Pattern so it can run in any environment Engine supports.
Also in the platform
Iris Natural-language Pattern generation · In TestingWe onboard in small cohorts as capacity allows. Request an invite and we'll be honest about whether your use case is one we can support today.