Investor Info · Datasets

The work behind
useful data.

Krummholz prepares openly licensed datasets for small businesses, applied AI teams and independent builders. Buyers pay for cleaned files, documented changes and an independent audit, with the original source license kept intact.

Preparing data takes attention: checking formats, handling duplicates, reviewing identifiers, building appropriate splits and keeping track of source terms. Our paid editions give buyers a documented starting point for that work.

Dataset progress · November 17, 2026Paid catalog open. Editions are released individually after an independent audit passes.
38editions listed across five focus areas
6editions available to buy in this dated snapshot
$36.80starting price per edition from November 18, 2026, before applicable tax
30 daysdownload access after a one-time payment

The catalog includes editions still in preparation. Listed editions are not all available products; check the live catalog for current availability.

Business model

Sell the preparation.
Keep the source rights intact.

Every paid edition links to its original publisher, where the source is free. The purchase covers Krummholz's curation, documentation and audit work. It does not buy exclusive ownership of the source data or add restrictions that conflict with its license.

Catalog editions

One listed price per edition, paid once through Stripe. Buyers receive download access for 30 days. There is no subscription or required sales call. Source licenses, cleaning reports, change logs, checksums and audit records travel with the files.

Custom dataset projects

Separately quoted work prepares a client's existing records for a defined task. Scope can include cleaning, labeling, format conversion and evaluation splits. Data permissions, confidentiality, review and delivery are agreed per project.

See what an edition includes · Read purchase terms · Explore custom data work

Recent progress

Six editions
ready to use.

The paid store opened on November 16, 2026. By November 17, six editions were available: TruthfulQA, SNIPS NLU Benchmark, eCFR Title 50, NOAA's Alaska groundfish survey, BAAD woody-plant measurements and Canada's National Forestry Database.

The store checks each edition's audit record before offering a buy button. Its product card shows the cleaning and audit figures available for that edition. Editions still being prepared show a planned price and cannot be purchased.

Visitors can now browse paid editions by subject, data type, license and price, with free releases in their own catalog.

An audit reduces known problems. Buyers still need to check whether a dataset fits their intended use.

Free releases and synthetic data

Room to try an idea.

Free downloads

Free releases give people a way to inspect records and try a workflow before buying. They live in a dedicated catalog with their own documentation, source terms and review status.

Browse free datasets

Synthetic data development

A free forestry test collection, its MIT-licensed generator and a QA preview show the work so far. A broader synthetic data program for model development and client projects remains in development, subject to source, privacy and quality review.

Inspect the synthetic releases and plans

What comes next

Expand by finishing the work.

The next step is to prepare and audit more of the listed editions, then make them available one at a time. Each release needs usable files, clear documentation and an audit record before it can be sold. Custom projects are scoped around a client's actual records and intended use.

Investor conversations

Talk with Hayven.

Krummholz AI is raising a venture round. To request the investor deck or discuss the dataset business, contact founder Hayven Jackson at info@krummholz.ai or (207) 307-0611.