01 / Ground truth
Nine canonical families
Civic, maritime, permits, workplace jobs, commuter flows, safety, POI, events, and urban signals arrive with one governed envelope.
Dataset product · Axiom Suite
AxiomCodex
The canonical intelligence dataset. Built for the work behind every decision.
270M+
rows & objects (est.)
279
registered data tables
218
populated tables
9
canonical families
Corpus figures estimated, Oct 2026; released packages are smaller.
The platform
A product story designed for the small screen.
Swipe through the system →
Chapter IIThe Catalog
Nine catalog datasets.
One normalization standard.
Every record ships with a schema spec, provenance record, and version history. Research tier is free. One All-access commercial license — $299/month or $2,990/year, plus applicable tax — covers every commercially licensed family.
LEHD Workplace Jobs (California 2023)
262,357 blocksComplete California 2023 Census LODES8 Workplace Area Characteristics (WAC): jobs per census block by worker age, earnings band, and NAICS sector. A free research release with no commercial rights; national 2023 and the 2002–2022 history are in Commercial All-access, which also includes one consolidated 2002–2023 collection (48,592,172 rows) delivered on request after purchase. Workplace jobs, not commuter flows.
Research releaseWorkplace jobsCensus blocksParquet
POI Intelligence
25,000-place sampleOverture Maps Places points of interest with Overture category taxonomy, H3 resolution-8 cell indexes, upstream confidence, and source observation timestamps. Commercial All-access includes packages for all 50 states + DC and one consolidated all-states collection (17,288,286 rows), delivered on request after purchase.
Source observationsCategorizedGeospatialParquet
Permit Signals
25,000-permit sampleA four-city research sample from Seattle, New York City, Chicago, and Los Angeles. Permit type, scope, status, and cost tiers use reviewed source mappings; building type and unit count are source-derived where available, not LLM labels. Commercial All-access separately covers the 424,996-row Seattle + Los Angeles snapshot, delivered on request after purchase.
Source recordsTemporalDevelopmentParquet
Civic Intelligence
517K+ civic recordsCouncil votes, permits, and zoning decisions structured with entity extraction, sentiment scores, upzoning probability, and DAG-mapped approval sequences.
NLP-readyLabeledTemporalParquetJSON Lines
Coming soonCrosswalk ↗
Not yet licensableAIS Maritime Positions
174M+ positions (est., Oct 2026)Decoded vessel tracks, port calls, and anchor events enriched with Equasis vessel metadata, flag state, DWT, and kinematic fingerprints.
Time-seriesEnrichedGeospatialParquet
Coming soonCrosswalk ↗
Not yet licensableUrban Signal Grid
H3 res-8 gridCell-level ESGI composite scores and 8 signal group subscores for U.S. metros, one row per H3 resolution-8 cell. Designed for CRE analysis and model workflows that need consistent H3-indexed urban signals.
GeospatialScoredMulti-signalGeoParquetCSV
Coming soonCrosswalk ↗
Not yet licensableAxiom Events Timeline
87M+ events (est., Oct 2026)Unified temporal intelligence — permits, council decisions, AIS anomalies, OSHA violations, business openings — normalized to a single event schema with source attribution.
TemporalMulti-sourceCross-domainJSON LinesParquet
Coming soonCrosswalk ↗
Not yet licensableLEHD Commuter Flows
Planned · no package yetPlanned: Census LEHD origin-destination commuter pairs aggregated to H3 cells, with income bands, job sector, and a Huff gravity index. No package has been built and this family is not licensable.
GeospatialDemographicsTransportCSVParquet
Coming soonCrosswalk ↗
Not yet licensableOSHA Safety Index
398K+ inspectionsOSHA inspection records with NLP-classified hazard categories, violation severity tiers, and inflation-adjusted penalty normalization by H3 cell.
NLP-classifiedSafetyIndustrialCSVParquet
Coming soonCrosswalk ↗
Not yet licensableFree research releases on Hugging Face
Public research downloads, no signup. Each dataset card states its upstream license and attribution; none grants commercial rights.
- LEHD Workplace Jobs (WAC) — California 2023 (opens Hugging Face in a new tab)
Every California workplace block with at least one job in Census LODES8 Workplace Area Characteristics for 2023 (262,357 blocks), with no sampling.
- POI Intelligence — California (Overture Places, full) (opens Hugging Face in a new tab)
All 1,941,553 eligible Overture Maps Places records assigned to California from release 2026-09-23.1, with upstream CDLA-Permissive-2.0 and Apache-2.0 terms and notices included.
- CDC PLACES Census Tract Health Estimates (2025) (opens Hugging Face in a new tab)
CDC PLACES 2025 modeled health estimates for 83,522 census tracts across all 50 states and DC; U.S. Government public domain with CDC attribution conditions.
- POI Intelligence — 25K sample (opens Hugging Face in a new tab)
A 25,000-row global sample of Overture Maps Places points of interest, normalized to the Codex POI schema.
- Permit Signals — 25K sample (opens Hugging Face in a new tab)
A 25,000-row research sample of building permits from the Chicago, Los Angeles, New York City, and Seattle open-data portals.
⬡
Schema-locked
Versioned schemas (semver). Breaking changes bump major. Consumers pin to a schema version and trust forward compatibility.
◈
Source-attributed
Every record carries source_feed, ingested_at, normalization_version, and confidence_score. No black-box data.
⌬
Documented source fields
Categories, taxonomies, and derived fields follow each package’s schema. Source observations are not adjudicated training labels; training requires separate governed authorization.
⊞
Documented spatial keys
Packages document their geographic grain and join keys, including H3 resolution 8 where supported. Align geography before joining.
⟳
Versioned snapshots
Versioned, dated snapshot releases with changelogs. Pin a release for reproducible analysis; permitted uses follow its source terms and approvals.
⊷
Cross-dataset joins
Use the shared keys each schema actually supplies — such as h3_index or jurisdiction_slug — and align grain and time coverage before joining.
How it fits
One data layer. Three products.
Locus and Overwatch are the application layer — intelligence tuned for specific markets. Codex is the data layer underneath, sold directly to builders who want the raw feed without the lens.
Application
Axiom Locus
CRE location intelligence. Scores, signals, and civic risk — interpreted for real estate decisions.
Application
Axiom Overwatch
Maritime intelligence. Vessel tracks, port congestion, AIS anomalies — for shipping and trade.
Dataset product
Axiom Codexyou are here
The same data — no lens. Canonical, schema-locked, AI-optimized. For builders who want the raw intelligence.
Licensing
Three tiers.
Research
Free
Source licenses apply
- ✓Public 25K-row samples plus full research releases
- ✓HuggingFace download
- ✓Community support
- ✓Schema spec access
Commercial
$299
/ month · All-access
or $2,990/year — 2 months free · plus applicable tax
- ✓Today: POI Intelligence, LEHD Workplace Jobs, Permit Signals (Seattle, Los Angeles)
- ✓Includes consolidated collections alongside the per-state and per-year packages: POI Intelligence for all 50 states + DC (17,288,286 rows) and LEHD Workplace Jobs 2002–2023 (48,592,172 rows)
- ✓New families added at no extra cost as they clear licensing
- ✓Versioned, dated snapshots; new releases as each family is refreshed
- ✓Package delivery on request by email after purchase
- ✓Schema changelog access
- ✓Email support
Enterprise
Custom
Talk to us about custom terms
- ✓Custom license terms, by conversation
- ✓Custom delivery arrangements
- ✓Custom schema extensions
- ✓Support beyond email
- ✓Commercially licensed families only; no redistribution today