Geofuchs

AI-powered geospatial analysis

ESA BIC Finland — March 25, 2026



Team & Experience

Jens Kreier — CEO/CTO

  • Spatial & Full-Stack Expert: Full-stack developer specialising in spatial data systems, advanced visualisation, and backend data pipelines.
  • Senior Consultancy: Senior Full-Stack Consultant at Rakettitiede — delivered R&D projects including an air quality dashboard for a major German car manufacturer in Asia.
  • Systems Architecture: Senior Software Architect with cloud-native AWS experience and multidimensional data UI tooling.
  • 3D & Advanced Visualizations: Highly experienced in building tools for multidimensional data and utilizing rendering frameworks like Unreal Engine and Deck.gl

Jarkko Nieminen — COO

  • B2B Integration & Scaling: Built digital data backbones for enterprise portfolios across Finland, Sweden, and Denmark.
  • Serial Entrepreneur: Track record of scaling digital startups from zero to market leadership, driving Go-To-Market strategy and operations.
  • Distribution Lead: Leads our major B2B consultancy partnership (200 consultants, 10 offices).

The Problem


Thousands of municipalities. One spatial question each. Nobody gets an answer today.


1

Fragmented Workflows

Work split across desktop GIS, scripts, email, and consultants — with limited reuse between projects.

2

Days Lost Per Report

Planning, environment, and resilience teams spend days combining datasets, running analyses, and packaging results.

3

Massive Scale

~92,000 local administrative units across the EU — small productivity gains compound fast.

Solution & Value Proposition

Ask in Finnish, Swedish or English → the system builds a reproducible GIS workflow with an interactive map layer, written explanation, and traceable sources and parameters — defendable, repeatable, auditable.

Days → Minutes

Manual work compressed dramatically for common spatial queries.

No GIS Expertise Required

Common queries answered without specialist knowledge.

Exports to Existing Tools

Not a rip-and-replace — can integrate with current GIS environments

AI Act Ready

Traceability, transparency, and human oversight built in by design

Space data is the backbone

Without it, the platform loses temporal analysis, change detection, and rural coverage.

Satellite data is architecturally essential. Without satellite access, Geofuchs loses change detection, environmental monitoring, and real-time event response — and becomes only a smarter map viewer.

Space Layer

Sentinel, Copernicus DEM

National Layer

MML, SYKE

Municipal Layer

Local datasets & planning data

Technology & IP

A lightning-fast, in-memory architecture bridging local spatial data with live Earth Observation.

Zero-Latency GIS Engine
Executes massive, multi-step spatial analyses entirely in-memory — no bottlenecks, no waiting.

AI-Driven Orchestration
Instantly translates natural language into optimised, self-healing analytical workflows.

Effective Binary Data Streaming
Streams complex spatial results to the 3D map at a fraction of traditional data transfer sizes.

Dynamic Satellite Routing
Automatically detects Earth Observation queries and offloads processing to live remote space backends.

Unified Data Fusion
Merges local authoritative data with real-time national APIs and satellite imagery in a single session.

Legal protection

Potential patent cases identified and IPR consultation on-going with dedicated legal advisers.

Trademark registration in progress.

Demo

Demo

Market Opportunity & Competitive Position

We complement — we don't compete

Public Sector Channel

Nordic entry wedge
92,000 local administrative units across the EU


B2B Channel

GIS consultancy market — established client networks

Reaches private sector organisations at scale

Revenue grows faster than headcount

Enterprise GIS

Built for specialists → We free the specialists to do higher-value work

Generic AI / LLM

No spatial execution → We execute and deliver results you can defend

Data-Only Competitors

Data without context → We connect spatial data to real-world decisions

Business Model: Dual revenue streams

Revenue Model

Public Sector Channel
Annual licensing

Tiered by organisation size

B2B Channel
Via major B2B GIS consultancy partnership
Scalable across enterprise partners

Financial Model

Gross Margin
~80%

Break-even
8–10 customers

Runway
12+ months on ESA BIC funding

Burn Discipline
Hiring gated by customer milestones

Risk Mitigation

Procurement Cycles
B2B partnership bypasses direct cold-selling; pilots validate ROI

Defensible Outputs
Every result traceable and explainable
User stays in control of every decision

Activity Proposal & Execution Milestones

Vision: Copernicus intelligence for every European municipality

Q2 '26: Sentinel Hub + openEO Integration

(25h ESA | €25K)
→ Production RS pipeline

Q3-Q4 '26: 3 Municipal Deployments

(15h ESA | €20K)
→ Measured time savings

Q4 '26: TorchGeo + SAR Change Detection

(30h ESA | €30K)
→ ESA benchmarked automation

Q1 '27: Accuracy Validation

(10h ESA | €15K)
→ Validated pipeline

Deliverables

3 active municipal deployments with measured ROI

Validated Earth Observation workflow module

Nordic GTM Pack: Ready for our B2B partners to scale

Thank You

Questions?


Annex I: Competitive Positioning Overview

No competitor scores across all 6 dimensions simultaneously. This is our defensible positioning gap.

Annex II: System Architecture (Part 1)

Geofuchs Hydra-8 is built on a robust, scalable architecture designed for high-performance geospatial intelligence.

Pipeline Engine

Single orchestrator converts natural language into executable geospatial workflows.

  • Qwen3.5-9B (xLAM merged base + LoRA v6, INT8) generates JSON DAG tool chains
  • DAG schema validation with automatic repair for malformed plans
  • DuckDB executor runs CTE-fused SQL chains per conversation session
  • 99 pipeline tools: vector, raster, routing, geocoding, Earth Observation

Semantic Resolution

Trilingual intent understanding (Finnish, Swedish, English) with multi-source context building.

  • Qdrant vector search across 1,144 indexed layers (similarity threshold 0.3)
  • Memgraph knowledge graph for entity relationships and spatial reasoning
  • Qwen2.5-1.5B SLM with RAG-grounded intent classification
  • Geographic guardrails reject non-geospatial or unresolvable queries

LLM Infrastructure

Self-hosted inference with fine-tuned models optimized for geospatial tool-calling.

  • vLLM v0.17.1 serving Qwen3.5-9B with BitsAndBytes INT8 quantization
  • LoRA adapter (v6) specialized for pipeline DAG generation
  • Multi-turn conversation manager with session-scoped DuckDB state
  • Arize Phoenix + OpenTelemetry for LLM observability and tracing

Annex II: System Architecture (Part 2)

Completing the Geofuchs Hydra-8 architecture with data and infrastructure layers.

Data Layer

Unified spatial querying across national providers and real-time feeds.

  • PostgreSQL/PostGIS: spatial data + Paavo postal statistics
  • TimescaleDB: time-series from Fintraffic TMS stations and road weather sensors
  • Redis: response and layer caching
  • WFS/WMS clients: live fetching from MML, SYKE, GTK, HSY, Digiroad, Fintraffic

Infrastructure

Kubernetes-native deployment with full observability.

  • k3s cluster with base + overlay manifests
  • FastAPI + Uvicorn API server with SSE streaming
  • Prometheus metrics, OpenTelemetry distributed tracing
  • Local container registry, versioned deploys via bump-my-version

Annex III: Data Coverage (Part 1)

Geofuchs Hydra-8 provides access to 1,144+ indexed geospatial layers across Finland and Sweden, delivering unparalleled data breadth for municipal intelligence.

At a Glance

Headline numbers for the entire data platform.

  • 1,144+ semantically indexed layers in Qdrant, 1,214 metadata entries, 307 routed datasources
  • 11M+ spatial features in PostGIS across 191 tables and 12 schemas
  • 7+ GB local PostGIS storage (6.5 GB MML topographic alone, 630 MB forest data)
  • 15 data providers (7 Finnish, 8 Swedish), 30 openEO satellite products, 3 satellite constellations
  • 3 routing tiers: PostGIS (<100ms), WFS (1-5s), openEO (satellite)

National Topographic & Cadastral

Complete Finnish land cover from the National Land Survey (MML).

  • 5.6M buildings, 3.8M road segments, 150K water bodies, 20K railway lines (PostGIS, EPSG:3067)
  • Cadastral boundaries, administrative regions, place names — 130 tables in the mtk schema alone
  • Metsäkeskus forest data: 860K forest stands, 244K forest mask polygons, METE habitats

Urban & Municipal

Detailed city-level data for the Helsinki Metropolitan Area.

  • Helsinki: buildings, addresses, green areas, tree registry, zoning, waterways, traffic volumes, public services (17 tables)
  • HSY: land cover, road network, regional buildings, vegetation (5 tables)
  • Espoo: tree inventory (4 tables)
  • LIPAS: national sports facilities, outdoor routes, recreation areas

Statistics & Demographics

Socioeconomic indicators from Statistics Finland.

  • 3,354 postal codes with 89 indicators: population, income, education, employment, housing (Paavo)
  • 1km population grid, municipal boundaries
  • Joinable with spatial geometries for thematic mapping

Annex III: Data Coverage (Part 2)

Completing Geofuchs Hydra-8 data coverage with real-time, satellite, and Swedish data sources.

Real-Time & Transport

Live sensor networks and infrastructure from Fintraffic, FMI, and Digiroad.

  • Traffic measurement stations (TMS), road weather sensors, traffic cameras, EV charging stations, railway stations.
  • Marine vessels via AIS tracking.
  • FMI: weather observations, road weather, marine data (WFS).
  • TimescaleDB ready for time-series polling when enabled

Earth Observation

Satellite imagery processed server-side via openEO (CDSE VITO backend).

  • Sentinel-2 (10m): 10 spectral indices — NDVI, NDWI, NBR, EVI, SAVI, and more.
  • Sentinel-1 (20m): SAR backscatter and coherence for flood and subsidence detection.
  • Sentinel-5P (1km): atmospheric trace gases — NO2, SO2, CO, O3, CH4, aerosol.
  • Temporal composites, change detection, cloud-filtered mosaics.
  • STAC catalog (Element84): Sentinel-2 + Landsat COG access with local caching

Swedish Coverage

8 national providers with hybrid PostGIS/WFS routing for major cities.

  • Lantmäteriet: properties, buildings, municipalities.
  • SGU: bedrock, soil types; Naturvårdsverket: nature reserves, national parks.
  • Skogsstyrelsen: forest land; SMHI: water bodies; SCB: population grid.
  • Local PostGIS for Stockholm, Gothenburg, Malmö — WFS fallback elsewhere