Training-data catalog v0.1

Browse AI-ready datasets with the license work already started.

Every listing is graded for license clarity, provenance, AI-readiness, and packaging status. Public/open datasets are sold as curation, access tooling, compliance docs, and premium licensed supply — not false ownership of open data.

100+
Visible records target
25
Priority cards live
42
Seed-approved
25
Use-case packs

License-first

No NC/ND datasets in production. Rights-complex sources stay metadata-only until reviewed.

Sample before full mirror

Large corpora ship as samples or metadata records first, then graduate to full R2 mirrors.

Use-case packaged

Each dataset maps to buyer jobs: RAG, fine-tuning, ASR, legal QA, coding, geospatial, and more.

Grade ASample readyAI-ready 94/100

Wikipedia English Articles Dump

Wikimedia Foundation · General knowledge · Text / encyclopedia · en

Source ↗
Size
~22 GB compressed / ~84 GB extracted
License
CC-BY-SA 4.0
Availability
10% processed sample first; full mirror queued after attribution manifest.
RAG starterKnowledge QAEntity extraction

Attribution: Requires Wikimedia attribution and share-alike obligations in downstream docs.

Grade ASample readyAI-ready 90/100

Project Gutenberg Public-Domain Books

Project Gutenberg · Literature / humanities · Text / books · en + multilingual

Source ↗
Size
~60,000+ books / ~5–10 GB metadata-first
License
Public Domain (US)
Availability
Curated biography/literature pack first; full corpus mirror in queue.
Writing AIBiography.ai corpusLong-context evaluation

Attribution: Public domain in the US; users must verify jurisdictional copyright status.

Grade AMetadata listedAI-ready 91/100

PubMed Open Access Biomedical Text

NIH / NCBI · Biomedical · Text / XML / abstracts · en

Source ↗
Size
~5 GB abstracts; OA full-text subset queued
License
CC-BY / CC0 / PD mix
Availability
License-per-article audit before production download bundle.
Biomedical RAGClinical NLPDrug-discovery literature

Attribution: Attribution and article-level license file required.

Caution: Article licenses vary; no undifferentiated bulk commercial pack until manifest is complete.

Grade BMetadata listedAI-ready 88/100

arXiv CS / AI / ML Papers

arXiv · Scientific research · Text / LaTeX / PDF · en

Source ↗
Size
~100–200 GB text-selected subset
License
Mixed open licenses + arXiv terms
Availability
Metadata and use-case pack listed; mirror requires license segmentation.
STEM RAGMath reasoningResearch assistant

Attribution: Per-paper license and arXiv attribution required.

Caution: Not every arXiv paper is CC-BY; segment before download sales.

Grade ASample readyAI-ready 92/100

CourtListener US Court Opinions

Free Law Project / CourtListener · Legal · Text / judicial opinions · en-US

Source ↗
Size
~100–350 GB depending subset
License
Public Domain + Free Law terms
Availability
Federal opinions sample first; full legal pack queued.
Legal QACitation extractionLaw.ai training/evals

Attribution: Free Law Project source attribution recommended.

Grade ASample readyAI-ready 93/100

SEC EDGAR 10-K / 10-Q / 8-K Filings

US Securities and Exchange Commission · Finance / corporate disclosures · Text / HTML / XML · en-US

Source ↗
Size
~200 GB selected recent filings; full archive multi-TB
License
Public Domain (US government)
Availability
Last five years of 10-K/10-Q selected pack queued for mirror.
Financial RAGRisk extractionCorporate.ai entity enrichment

Attribution: SEC source attribution and fair-access notes included.

Grade ASample readyAI-ready 89/100

Wikidata Entity Graph Dump

Wikimedia Foundation · Knowledge graph · Structured / JSON graph · multilingual

Source ↗
Size
~115 GB compressed
License
CC0
Availability
Entity subset sample ready; full dump queued.
Entity resolutionKnowledge graph RAGCorporate.ai enrichment

Attribution: CC0; source provenance retained for buyer trust.

Grade BMetadata listedAI-ready 84/100

OpenStreetMap Planet Data

OpenStreetMap contributors · Geospatial · Structured / map graph · global

Source ↗
Size
~70+ GB compressed planet file
License
ODbL 1.0
Availability
Metadata and regional sample plan listed; full mirror after ODbL notice pack.
RoutingPOI extractionGeospatial forecasting

Attribution: ODbL attribution/share-alike obligations must be surfaced.

Grade AMetadata listedAI-ready 90/100

Mozilla Common Voice 17

Mozilla Common Voice · Speech · Audio / transcripts · multilingual

Source ↗
Size
3,000+ hours multilingual
License
CC0
Availability
Language-priority samples queued: EN, ES, FR, DE, HI.
ASRAccent robustnessSpeech evals

Attribution: CC0; retain source provenance.

Grade AIngestion queuedAI-ready 91/100

LibriSpeech train-clean/train-other 960h

OpenSLR · Speech · Audio / read speech · en

Source ↗
Size
~960 hours
License
CC-BY 4.0
Availability
Queued as first full audio mirror candidate.
ASR baselineSpeech encoder trainingAudio QA

Attribution: LibriVox/OpenSLR attribution required.

Grade AIngestion queuedAI-ready 87/100

CodeSearchNet Code + Docstrings

GitHub / CodeSearchNet · Code · Code / text pairs · multiple programming languages

Source ↗
Size
~20 GB
License
MIT
Availability
Queued for coding-assistant starter pack.
Code searchCoding assistant fine-tuneDocstring generation

Attribution: MIT license notice retained.

Grade BMetadata listedAI-ready 86/100

The Stack — Permissive License Subset

BigCode / Hugging Face · Code · Code repositories · programming languages

Source ↗
Size
Multi-TB; permissive subset first
License
MIT / Apache / BSD subset
Availability
Metadata-only until permissive-license filter and opt-out compliance are documented.
Code LLM pretrainingLanguage-specific packsRepo QA

Attribution: License-per-file manifest required.

Caution: Only permissive-license subset is production eligible.

Grade AIngestion queuedAI-ready 88/100

GovInfo Federal Register + CFR Bulk Data

US Government Publishing Office · Government / regulatory · Text / XML · en-US

Source ↗
Size
~20+ GB selected regulatory text
License
Public Domain
Availability
Queued for compliance and civic-data RAG packs.
Regulatory RAGVote.net civic packsCompliance QA

Attribution: US government source attribution retained.

Grade AIngestion queuedAI-ready 85/100

USPTO Patent Grant Text

USPTO Bulk Data · Patent / IP · Text / XML · en-US

Source ↗
Size
~150 GB compressed selected corpus
License
Public Domain
Availability
Queued after SEC/legal priority packs.
Patent searchR&D analyticsIP claim extraction

Attribution: USPTO source retained.

Grade BIngestion queuedAI-ready 82/100

UCI Machine Learning Classics Pack

UCI ML Repository · Tabular ML · CSV / tabular · various

Source ↗
Size
Small curated bundle
License
Mixed open / citation required
Availability
Queued as Explorer-friendly tabular starter bundle.
Tabular classificationAutoML demosTeaching datasets

Attribution: Dataset-specific citations required.

Grade AMetadata listedAI-ready 80/100

NASA Earth Observations / Public Science Data

NASA Open Data · Climate / space / earth science · Structured / imagery metadata · global

Source ↗
Size
Dataset-specific; metadata-first
License
Public Domain (US government)
Availability
Metadata-only while pack boundaries are selected.
Climate forecastingRemote sensingScientific RAG

Attribution: NASA source attribution retained.

Grade AMetadata listedAI-ready 83/100

NOAA Weather and Climate Open Data

NOAA · Weather / climate · Structured / time series · US + global

Source ↗
Size
Large; selected time-series pack first
License
Public Domain (US government)
Availability
Metadata-only pending first regional/time-window sample.
ForecastingRisk modelingClimate analytics

Attribution: NOAA source attribution retained.

Grade AIngestion queuedAI-ready 86/100

FEC Campaign Finance Data

Federal Election Commission · Civic / elections · Structured / CSV · en-US

Source ↗
Size
~GB-scale bulk data
License
Public Domain
Availability
Queued for Vote.net civic-data pack.
Election analyticsCivic RAGEntity matching

Attribution: FEC source retained.

Grade AIngestion queuedAI-ready 85/100

USAspending Federal Awards Data

USAspending.gov · Government spending · Structured / CSV · en-US

Source ↗
Size
Large; selected fiscal-year packs
License
Public Domain
Availability
Queued for procurement and corporate intelligence packs.
Vendor intelligenceProcurement RAGGovernment spend analytics

Attribution: USAspending source retained.

Grade BMetadata listedAI-ready 78/100

Open Food Facts Product Database

Open Food Facts · Food / nutrition · Structured / images metadata · global

Source ↗
Size
~GB-scale exports
License
ODbL 1.0 / images mixed
Availability
Metadata-only until image license split is complete.
Nutrition AIProduct matchingFood labeling QA

Attribution: ODbL attribution/share-alike obligations required.

Caution: Images may carry separate licenses; split text/structured first.

Grade AMetadata listedAI-ready 87/100

OpenAlex Scholarly Metadata Graph

OpenAlex · Scholarly metadata · Structured / graph · global

Source ↗
Size
Hundreds of GB snapshot
License
CC0
Availability
Metadata-only pending snapshot partition plan.
Research graphCitation RAGExpert discovery

Attribution: CC0; provenance retained.

Grade BIngestion queuedAI-ready 84/100

ChEMBL Compound and Bioactivity Data

EMBL-EBI ChEMBL · Life sciences / chemistry · Structured / SQL / TSV · global

Source ↗
Size
~GB-scale database dumps
License
CC-BY-SA 3.0
Availability
Queued for drug-discovery starter pack with share-alike warning.
Drug discoveryBioactivity modelingMolecular RAG

Attribution: ChEMBL attribution and share-alike obligations required.

Grade AIngestion queuedAI-ready 86/100

UniProt Knowledgebase

UniProt Consortium · Life sciences / proteins · Structured / FASTA / XML · global

Source ↗
Size
~100+ GB depending format
License
CC-BY 4.0
Availability
Queued after ChEMBL/PubMed pack definition.
Protein annotationBiomedical RAGBiosequence models

Attribution: UniProt attribution required.

Grade BMetadata listedAI-ready 82/100

Open Images Dataset Metadata + Annotations

Google Open Images · Computer vision · Images / annotations · global

Source ↗
Size
Annotations first; images URL-based
License
CC-BY 4.0 annotations; images mixed
Availability
Annotations/metadata listed; no image mirroring until image-level license check.
Object detectionImage captioningVision evals

Attribution: Image-level author/license attribution required.

Caution: Do not mirror underlying images without per-image license validation.

Grade CMetadata listedAI-ready 76/100

Common Crawl WET Text Sample

Common Crawl Foundation · General web text · Text / WET · multilingual

Source ↗
Size
1–5 TB initial monthly-crawl subset
License
Common Crawl terms / public web
Availability
Metadata-only until commercial-use caution language is reviewed.
Pretraining sampleWeb text filteringLanguage ID

Attribution: Common Crawl source attribution; downstream rights vary by page.

Caution: High-value but rights-complex; sell filtering/compliance tooling, not blanket clearance.