STEP 01
Ingest spatial evidence
Drone, satellite, lidar, and rasters — plus vectors, GeoJSON, CSV, and field data, unified in one place.
SatelliteDroneLiDARVectorsGeoJSONCSV
NIKA Planet
Geospatial AI & ML Workbench that Scales
Work with aerial images of every kind, from drone and satellite scenes to lidar, point clouds, rasters, vectors, and field evidence. NIKA Planet gives teams one cloud platform to complete the annotate-train-deploy loop and turn model outputs into maps, dashboards, and reusable spatial apps.
The platform
One workspace for geospatial AI and ML.
Maps, notebooks, labeling, dashboards, and shared data — every workflow runs in one place, on top of your data lake.
Cloud platform layers
How teams use NIKA Planet to deploy ML models end to end.
From raw aerial evidence to deployed map intelligence — one platform for the whole loop. Scroll to follow the path.
STEP 02
Annotate & assign
SAM-assisted masks with team assignment — route each task to two annotators who must agree before it is marked ready for review.
STEP 03
Clean & convert to cloud-native
Clean and preprocess every layer, then convert to cloud-optimized formats like COG and Zarr for fast, scalable access.
STEP 04
Train models in parallel
Run notebooks and GPU jobs next to the data — and fan out many runs at once with NIKA's native parallel Runner SDK.
STEP 05
Evaluate & validate
Compare accuracy, IoU, and experiments side by side, tracked end to end with MLflow.
MLflow3 runs comparedAccuracy94%IoU0.87F10.91
STEP 06
Deploy as inference functions
Ship each model as an inference function — on the map, via API, or straight into enterprise GIS like ArcGIS Pro and QGIS.
STEP 07
Version, retrain & redeploy
Every deployment is versioned, so models retrain on new evidence and redeploy in one click — closing the loop back to labeling.
Feature highlights
Every module of the annotate-train-deploy loop, explained.
Workspace, Map, Labeler, Notebook, Flow, DBMS, Dashboard, Hub, and GeoEngine — each a focused product, all on one cloud operating layer for geospatial AI. Here is what each one does and the outcomes it delivers.
NIKA Workspace
One central home for all your spatial data and AI analysis.
Workspace is a petabyte-scale geospatial data lake where teams store, process, and analyze massive datasets in one shared place. Analysts ask questions in plain English and let the built-in AI surface patterns and anomalies, while folders auto-sync between the web and connected machines — one source of truth instead of scattered files and tools.
Key traits
- Petabyte-scale storage
- Plain-English AI queries
- Real-time filesystem sync
NIKA Map
Desktop-class mapping for huge datasets, right in the browser.
Map is a GPU-accelerated, WebGL mapping framework that renders millions of features and gigabytes of imagery fluidly — nothing to install. Teams edit the same map together with Google-Docs-style auto-save, live cursors, and full version history, so decisions get made faster with a clear audit trail of who changed what.
Key traits
- GPU-accelerated rendering
- Real-time co-editing
- Version history & audit trail
NIKA Labeler
Annotate satellite and drone imagery together, faster, with AI.
Labeler is a geospatial-first annotation platform built for top-down imagery, with many people labeling at once in real time. AI detection and text-prompted segmentation let users tag objects just by describing them, while a built-in reviewer approval flow keeps quality high — labeled datasets ship faster with consistent oversight.
Key traits
- Real-time co-annotation
- AI text-prompted segmentation
- Built-in approval workflow
NIKA Notebook
Collaborative Python notebooks built for spatial analysis at scale.
Notebook is a Jupyter environment purpose-built for geospatial work, with geo libraries preinstalled and real-time co-editing. Teams pick the compute they need — from CPU-only up to GPU machines — and heavy jobs keep running server-side even if a connection drops, removing infrastructure setup friction.
Key traits
- Real-time collaboration
- Flexible CPU / GPU compute
- Jobs run even offline
NIKA Flow
Convert geospatial and 3D data between formats.
Flow is a cloud ETL tool focused on data conversion — translating geospatial and 3D datasets from one format to another, and reprojecting coordinate systems along the way. Teams use it to standardize incoming data and move it between tools without writing one-off conversion scripts.
Key traits
- Format-to-format conversion
- Geospatial & 3D formats
- Coordinate reprojection
NIKA DBMS
Ask your spatial databases questions in plain English.
DBMS turns natural-language requests into optimized spatial SQL by understanding your schema. It connects across PostgreSQL, DuckDB, Snowflake, BigQuery, and more, and queries Workspace data-lake files directly without duplication — answers come faster across mixed database infrastructure, with query history kept for reproducibility.
Key traits
- Natural language to SQL
- Multi-database support
- Direct data-lake access
NIKA Dashboard
Live, location-aware dashboards built without writing code.
Dashboard is a real-time analytics builder for geospatial business intelligence, assembled through a drag-and-drop interface with live preview. It connects to APIs, databases, CSVs, and IoT streams, integrates natively with NIKA Map, and supports full branding and permissions — decision-ready dashboards without development resources.
Key traits
- Drag-and-drop builder
- Live multi-source data
- Native map integration
NIKA Hub
Discover, share, and publish geospatial apps and analyses.
Hub is where teams discover, share, and showcase geospatial apps, maps, dashboards, and notebooks. One-click publishing with public, team-only, or password-protected access lets organizations reuse ready-made templates and industry workflows — or share their own — so teams start from proven examples instead of reinventing the wheel.
Key traits
- One-click publishing
- Granular access controls
- Industry solution library
NIKA GeoEngine
Reproducible, GPU-ready geospatial compute inside your GIS tools.
GeoEngine packages Python and R geospatial workloads into reproducible, GPU-accelerated Docker containers with a single command. Semantic versioning lets teams rerun the exact same algorithm months later, and workers appear natively inside ArcGIS Pro and QGIS toolboxes — dependable, repeatable results, with AI agents able to scaffold and deploy workers automatically.
Key traits
- Absolute reproducibility
- GPU acceleration
- Native ArcGIS / QGIS integration
NIKA Runner SDK
Fan Python jobs out across many cloud CPUs and GPUs in parallel.
Runner SDK delegates Python to the cloud with a single @run decorator — pick the machine (CPUx20, T4x1, H100x1), then submit jobs that run in parallel across many runners at once. An orchestrator run spawns child runs that fan across CPU and GPU compute together, and every run is tracked live with status, logs, and a run graph.
Key traits
- One @run decorator
- Parallel CPU & GPU runners
- Live run-graph tracking
Why NIKA Planet
What makes NIKA Planet different for ML scientists.
General-purpose ML and data platforms make you rebuild the geospatial stack yourself — projections, aerial imagery, spatial SQL, the labeling-to-deployment pipeline. NIKA Planet makes all of it native, on one petabyte-scale spatial data lake, so geospatial ML scientists ship models instead of plumbing.
Environment setup
The old way
Wrestle GDAL, PROJ, and CUDA into a working build — fighting binary conflicts before a single line of geospatial code runs.
With NIKA Planet
Notebooks ship with GDAL and CUDA preinstalled and spatial libraries native across every layer, so you start analyzing immediately.
GDAL + CUDA preinstalledAerial & drone imagery
The old way
Satellite and drone rasters are treated as plain RGB — the georeferencing and coordinate systems your work depends on get quietly dropped.
With NIKA Planet
Top-down imagery is first-class; coordinate systems are auto-detected and preserved from ingestion through labeling and training.
Auto CRS · GeoTIFF / COGQuerying your data
The old way
A generic text-to-SQL assistant has no idea what a spatial join, a projection, or a spatial index even is.
With NIKA Planet
Schema-aware plain English becomes optimized spatial SQL — ST_* functions and spatial joins — across your connected engines.
NL → spatial SQL · PostGIS · DuckDBRendering results
The old way
Static plots and BI tiles choke the moment a spatial layer gets large, so you down-sample and lose the detail that matters.
With NIKA Planet
WebGL renders millions of features and gigabytes of raster live in the browser — pan, zoom, and inspect at full resolution.
Millions of features · in-browserAnnotate → train → deploy
The old way
A separate labeling tool, a separate notebook, and a separate MLOps stack — stitched together by hand-moving datasets between them.
With NIKA Planet
One loop: AI-assisted labeling on the same data lake your notebooks train on, exported COCO-ready for the model step.
SAM2 / SAM3 · COCO exportShipping to GIS users
The old way
Wrap every model in a bespoke API or proxy service just so analysts in their desktop GIS can finally use it.
With NIKA Planet
GeoEngine publishes a version-pinned, GPU-ready container that appears as a native tool inside ArcGIS Pro and QGIS.
Reproducible · ArcGIS / QGISThe difference is the same one across every row: general-purpose platforms make you rebuild the geospatial stack yourself. NIKA Planet makes it native — so you ship models, not plumbing.
Get started
Two ways to run NIKA Planet.
NIKA Planet runs as a fully managed cloud platform you can trial today — or deploy it inside your own environment when data residency, security, or compliance require it.
Cloud teams
Start in our cloud, today.
NIKA Planet is already hosted and managed in our cloud. Spin up a workspace and run the full annotate-train-deploy loop in minutes — no infrastructure to stand up, always on the latest version.
- Managed & automatically updated
- Nothing to install or host
- Scale GPU compute on demand
On-prem & self-hosted
Deploy in your own environment.
Run NIKA Planet inside your own cloud, data center, or air-gapped network when data residency, security, or compliance require it. Our team scopes the deployment and sizing with you.
- Your VPC, data center, or air-gapped
- Data never leaves your environment
- Tailored to your compliance needs