Waiheke Data

Eastern European data you can't scrape.

Licensed, anonymised and structured datasets for training and evaluating AI models. Romanian and Moldovan sources, built and processed in the EU.

Request a sample
RomaniaMoldovaEU processed

Datasets

What we supply

AI training data from distinctive sources, prepared for model training and evaluation.

02

Receipts and invoices

A document AI dataset of real Romanian fiscal receipts and invoices for OCR: images plus extracted fields, with personal data removed.

03

Historical texts

Public-domain Eastern European books, cleared by the author's date of death. OCR text with pre-1953 Romanian spelling and the Cyrillic-to-Latin transition.

04

Knowledge points

Processed extracts from our sources: facts, Q&A pairs, entity relations and reasoning chains. No personal data, so they ship to any jurisdiction.

05

Anonymisation pipeline

Data anonymisation for Romanian legal and administrative text, offered as a service or installed on your own servers so documents never leave your premises.

06

Romanian speech

Recorded talks and conferences in Romanian, only where speakers have signed rights for AI training. Limited volume, growing.

Clear before delivery

Every dataset ships with

  • A data card and schema
  • Documented provenance and rights
  • Anonymisation method and residual-risk report
  • JSONL or Parquet delivery
  • A licence with no redistribution

Inspect the structure

What a data card looks like

Example record (illustrative)

JSONL
record_id
RO-CIV-000123
jurisdiction
Romania
court_level
Court of Appeal
area_of_law
Civil – contracts
year
2023
outcome
Appeal admitted, first-instance ruling reversed
legal_basis
Civil Code art. 1350
parties
[PARTY_A], [PARTY_B] (anonymised)
reasoning_steps
  1. facts
  2. legal issue
  3. applied rule
  4. application
  5. conclusion
paired_with
first-instance decision RO-CIV-000122

Every identifier is replaced before delivery. Each dataset reports its measured residual risk.

From first look to licence

How we work

  1. 1

    Free sample

    20–50 records you can inspect

  2. 2

    Spec and data card

    schema, volumes and rights, agreed in writing

  3. 3

    Paid pilot

    a first batch against your acceptance criteria

  4. 4

    Licence or subscription

    one-off or recurring delivery

Straight answers

Frequently asked

Where does the data come from?

Public court decisions, public-domain books, documents contributed with written consent by the companies that issued them, and recordings with signed rights. We document provenance for every dataset and share it under NDA.

Is it GDPR compliant?

We are an EU company and process all data in the EU. Personal data is removed before delivery; datasets that still contain personal data are only shared under EU standard contractual clauses.

Can we commission a custom dataset?

Yes. Tell us the domain, language, volume and format, and we will scope it with you.

What licence do you offer?

Non-exclusive licences for training and evaluation, with no redistribution. Exclusive terms are possible for custom work.

Start with a sample

Request a sample

Tell us which Romanian dataset you need and what you are evaluating.

Data type of interest *

About Waiheke AI

Waiheke AI is a Romanian AI company. Our education platform ZVÂC serves 11,000+ students across Africa and partner universities in Romania, the UAE and Europe.

Robert Csergo

Co-founder & CEO

robert@waiheke.ai+40 742 521 785