Intelligence begins from orbit.
The satellite data exists. We built the layer that makes it mean something.
MODIS
SENTINEL-2 L2A
ESA CCI PERMAFROST
SENTINEL-1 SAR
LANDSAT 8/9
ESA CCI BIOMASS
NASA GIBS
PIXXEL HYPERSPECTRAL
PIXXEL HYPERSPECTRAL
PIXXEL HYPERSPECTRAL
001
who we are
InkySalt Technologies is the only Canadian deep tech platform focused on Critical Minerals, Arctic Infrastructure, and Carbon Intelligence at once — one sovereign AI architecture that's always learning. The data was never the bottleneck. The bottleneck was the missing AI layer that turns it into regulatory-ready intelligence — for the explorer deciding where to drill, the operator watching a road on thawing permafrost, and the developer trying to prove a carbon claim. STRATUM, ARCTA, and CarbonScope are those layers: Three domains most companies treat as separate industries. We treat them as one architecture.
002
What we're solving
One pipeline. Three domains.
Satellite acquisition and AI synthesis never change. What sits between them is domain-specific and interchangeable, because the architecture was built that way from the first line of code.
01
MINERAL EXPLORATION
SENTINEL-2 L2A
SENTINEL-1 SAR
CLOUD COVER FILTER
MINERAL-SPECIFIC SPECTRAL INDEX
Z-SCORE ANOMALY NORMALIZATION
ADAPTIVE DBSCAN CLUSTERING
RAG INTELLIGENCE LAYER
CHROMA DB VECTOR DATABASE
NDVI VEGETATION MASKING
FAULT & SHEAR-ZONE LINEAMENTS
FOUR-FACTOR CONFIDENCE HIVEMIND
COMPLIANCE OVERLAY
COSINE SEARCH OVER 175 YEARS OF GSC DATA
v1 · in development
PIXXEL HYPERSPECTRAL · 400+ BANDS AT 5M
02
ARCTIC INTELLIGENCE
ARCTA Engine
SENTINEL-1 INSAR
RADARSAT CONSTELLATION
SENTINEL-2 CHANGE DETECTION
ESA CCI PERMAFROST HEATMAP
ARCTIC THAW MAPPING
CORPUS GROUNDING
RISK-SCORING SCHEMA
v1 · in development
LIVE PERMAFROST SUBSIDENCE
CROSS-ENGINE SYNTHESIS
ANOMALOUS ROUTING DETECTION
DARK VESSEL MONITORING
INFRASTRUCTURE INTELLIGENCE
30 YEARS OF HISTORICAL TIMELAPSE
CROSS-DOMAIN COMMAND INTERFACE
03
CARBON MONITORING
CARBONSCOPE Engine
v1 · in development
LANDSAT 8/9
NASA GEDI LIDAR
ESA CCI BIOMASS DENSITY
BIOMASS INDEX COMPUTATION
CARBON STOCK ESTIMATION
LAND COVER CHANGE DETECTION
RAG INTELLIGENCE LAYER
LIVE BIOMASS CHANGE DETECTION
PEATLAND EXTENT
CARBON STOCK MONITORING
FULL MRV REPORT GENERATION
Every layer is stress-tested
against real ground truth
before it ships.
BUILD → TEST AGAINST GROUND TRUTH → VALIDATE → SHIP
DATA PIPELINE IS PASSED THROUGH A RAG INTELLIGENCE LAYER — THE
DISCIPLINE CROSS-REFERENCING EVERY SATELLITE-DETECTED FINDING
AGAINST A DOMAIN-SPECIFIC CANADIAN CORPUS, WHERE EACH ENGINE
RETRIEVES FROM ITS OWN CORPUS:
STRATUM — 175 YEARS OF GSC OPEN DATA
ARCTA — CANADIAN PERMAFROST SCIENCE, TRANSPORT CANADA AND
TERRITORIAL ENGINEERING STANDARDS, HISTORICAL INSAR BASELINE RECORDS
CARBONSCOPE — ECCC CARBON OFFSET PROTOCOLS, IPCC MRV GUIDELINES,
PROVINCIAL ENVIRONMENTAL ASSESSMENT STANDARDS
003
process
How It Works
The satellite data was always there. The intelligence wasn’t.
The roadmap isn’t a wishlist. It’s the next engineering milestone.
01
Satellite Acquisition
Sentinel-2, Sentinel-1, Landsat, GEDI — Google Earth Engine pulls the most recent cloud-free imagery for your area of interest directly from the archive. No local download, no GIS software required
Domain-Specific Detection
Spectral mineral indices for STRATUM, ground displacement or optical change for ARCTA, biomass and land-cover change for CarbonScope — the one layer that's genuinely different per engine, running on a shared satellite-to-intelligence foundation.
02
03
Corpus-Grounded Retrieval
Every detected anomaly is cross-referenced against a domain-specific Canadian corpus — 175 years of GSC geology, engineering standards, or ECCC carbon protocols — through semantic vector search. Not a generic AI guess; a finding anchored to published science.
Confidence-Scored Synthesis
A narrative brief is generated and scored across multiple independent factors — spectral intensity, spatial co-occurrence, compactness, scale — not a single opaque number.
04
05
Regulatory-Ready Delivery
NI 43-101 remote sensing section, engineering risk brief, or MRV-formatted carbon report — output arrives already shaped for the framework you actually operate under.
004
Sovereign data sources
Built on Open Data,
Not proprietary lock-in
Sentinel-2
Landsat 8/9
GEDI Lidar
Sentinel-1
ESA CCI
Pixxel Firefly
Earth Engine
Sentinel-2
Landsat 8/9
GEDI Lidar
Sentinel-1
ESA CCI
Pixxel Firefly
Earth Engine
Sentinel-2
Landsat 8/9
GEDI Lidar
Sentinel-1
ESA CCI
Pixxel Firefly
Earth Engine
InkySalt runs on the same open satellite and scientific infrastructure operated by ESA, NASA, and the Canadian government — not a closed pipeline we control access to
005
Engines at work
Layers In Action
Retrieval-Grounded Narrative Synthesis
Every finding cross-referenced against a domain-specific corpus via semantic vector search before a word of narrative gets written
Multi-Index Spectral Analysis
Eight independent band-ratio indices, each tuned to a specific mineral system's surface alteration signature. Pixxel Hyperspectral would take this to the next level.
ARCTA — Testing
RADARSAT-2 Sovereign Coverage
Canadian SAR constellation integration for higher-latitude coverage, opening a federal/defence procurement pathway
Multi-Temporal Persistence Filtering
Comparing seasonal satellite composites to separate stable rock exposure from transient noise like tilled fields or snowmelt
Multi-Index Spectral Analysis
Eight independent band-ratio indices, each tuned to a specific mineral system's surface alteration signature. Pixxel Hyperspectral would take this to the next level.
Adaptive Spatial Clustering
DBSCAN with an epsilon derived per-scene from a k-distance graph, not a fixed value — the clustering scale adapts to how dense the actual anomalies are
Multi-Layer Environmental Masking
Vegetation, water, and land-cover exclusion applied before statistical baselining, so a river or a cropland field can't silently skew what counts as anomalous
Permafrost Reference Layer
ESA CCI Permafrost open data integrated as ground-temperature and active-layer-thickness context
Vegetation Stress Detection
Sentinel-2 NDVI anomaly mapping catching thermokarst symptoms — retrogressive thaw slumps, vegetation die-off — ahead of any subsidence measurement
Millimetre-Scale SAR Interferometry
Sentinel-1 phase-unwrapping for ground displacement detection, ARCTA's next engineering milestone
CarbonScope - In Development
MRV-Grade Reporting
Carbon accounting structured to IPCC MRV guidelines and ECCC offset protocols, not a generic emissions estimate
That's 9 new ones on top of the 8 STRATUM cards from before
Thermal Proxy Analysis
MODIS land surface temperature anomaly detection as a permafrost degradation proxy, daily composites at 500m resolution
STRATUM - v0.9
Four-Factor Confidence Scoring
Spectral intensity, spatial co-occurrence, compactness, and scale combined into one transparent score — never a single opaque number
Hyperspectral Confirmation Layer
400+ band imagery as a drill-ready confirmation pass over our highest-confidence targets
Biomass Density Mapping
GEDI spaceborne lidar fused with ESA CCI Biomass to estimate carbon stock per hectare with zero field plots required
Land Cover Change Detection
Landsat 8/9 multi-date comparison catching deforestation, regrowth, and peatland disturbance
Multi-Temporal InSAR Stack
Persistent scatterer interferometry across a 3-5 year time series, separating seasonal freeze-thaw from genuine long-term subsidence
006
FAQs
Questions That Matter
1
Is InkySalt's technology actually operational, or is this a concept?
STRATUM, our mineral prospectivity engine, is live and operational — it's processed real satellite data across six Canadian provinces and ARCTA is in beta rollout phase. CarbonScope a different underlying architecture but is earlier in development. Ask SIGNAL — Cross-Domain Command Interface coming soon.
2
How is this different from just buying satellite imagery?
Satellite data providers sell you the picture. We sell you what it means. STRATUM cross-references every detected anomaly against 175 years of Geological Survey of Canada literature before generating a single word of narrative — a geologist reviewing our output gets a defensible interpretation grounded in published science, not a raw pixel map they still have to interpret themselves
3
What makes this "sovereign" or Canadian-specific?
Every satellite we use — Sentinel-1, Sentinel-2, Landsat, GEDI — is open, international data, not proprietary or foreign-controlled. Every corpus we retrieve against GSC geology, Canadian engineering standards, ECCC carbon protocols — is Canadian. No foreign infrastructure sits between the data and the decision. For government and institutional buyers operating under Canadian data governance requirements, that's not a nice to have, it's often a procurement requirement.
4
Who is this actually for?
Three different buyers, one platform. Junior miners who can't afford a $900M exploration program get STRATUM. Northern governments watching permafrost eat their roads get ARCTA. Carbon developers priced out of MRV verification get CarbonScope. Same architecture, three domains — that's the platform, not three separate products.
5
Who's actually building this?
A solo founder, no funding and long nights protoyping a system designed to get sharper the more it's tested. Every layer runs through the same loop — build it, stress-test it against real ground truth, fix what breaks, ship the improved version. Test and repeat.
6
What makes STRATUM different from other exploration tools?
Other tools show you imagery — you still need a geoscientist to interpret it. STRATUM fuses satellite detection with 175 years of GSC literature to return a ranked, regulatory-ready target list in under 15 minutes, at a fraction of the $150K–$500K a conventional first-pass program costs. As we add hyperspectral and structural layers, the goal is shifting the industry’s starting point from “where might we look” to “where should we drill.”
7
How is ARCTA different from current Arctic monitoring?
Most corridors get checked once a year by helicopter, at $25K–$80K a trip. ARCTA replaces that with continuous satellite monitoring — our optical change-detection engine already independently flagged a real, documented thaw slump within ~145 metres, unprompted. Full Sentinel-1 InSAR, our next milestone, brings that to millimetre-scale precision. With 250,000km of infrastructure on degrading permafrost, catching failure before it happens instead of after is the whole point.
8
What will CarbonScope do, and when is it coming?
CarbonScope is our third domain, in development, applying the same architecture to carbon accounting — GEDI lidar and biomass data fused with MRV literature. Third-party carbon verification costs $50K–$200K per project today, pricing out most developers. CarbonScope is built to remove that bottleneck as Canada’s carbon market matures.
Have any other questions?
View on Desktop
View on Desktop