Bank and card transactions
10,000 customers
Synthetic retail bank and card activity: customers, accounts, cards, merchants, and payments.
txn_id · ts_utc · account_id · rail · direction · amount · currency · auth_result
CSV · Parquet · JSONL
SchemaAI solutions
We help teams build, train, and evaluate AI.
Synthetic datasets, training environments, and custom data work.
01/Work
01
Bank and card transactions, fraud and AML labels, and healthcare encounters and claims. Diagnoses use ICD-10-CM. Procedure codes are coming later.
02
Environments and graded task sets for training and evaluating AI agents. Company in a box is the current environment: a synthetic company with a month-end close.
03
Custom data and AI work, scoped to a request. Public-record data feeds are part of this work. DOT asphalt bid data is one example.
Product 01
Bank and card transactions, fraud and AML labels, and healthcare encounters and claims. Diagnoses use ICD-10-CM. Procedure codes are coming later. Listed sizes are a launch specification, not a count of rows sold. These sets are sold on Datarade. This site does not take payment.
01
Synthetic retail customers, deposit and card accounts, and the payments those accounts make.
10,000 customers
Synthetic retail bank and card activity: customers, accounts, cards, merchants, and payments.
txn_id · ts_utc · account_id · rail · direction · amount · currency · auth_result
CSV · Parquet · JSONL
Schema02
Fraud and AML labels written by the same bank simulation, for model training and testing.
10,000 customer sample
Labeled bank and card transactions, including a flat view for training fraud and AML models.
txn_id · is_fraud · fraud_type · is_aml · aml_typology · difficulty · scenario_id
CSV · Parquet · JSONL
Schema03
Synthetic patients, encounters, and ICD-10-CM diagnoses. Procedure codes are coming later. No real patients.
200 example encounters
Synthetic patients, encounters, ICD-10-CM diagnoses, and claims. Procedure codes are coming later.
patient_id · encounter_id · encounter_type · encounter_date · primary_icd10 · primary_icd10_description · claim_status · billed_amount_usd
CSV · Parquet · JSONL
SchemaAlso available
Multi-turn customer and agent chats for training and evaluating support models.
Instruction, input, and output examples with rubrics, split for training and evaluation.
Product 02
Environments and graded task sets for training and evaluating AI agents. Company in a box is the current environment: a synthetic company with a month-end close.
Environment
A fully synthetic, internally consistent company for training and evaluating AI agents.
v1 is a US outpatient clinic back office, with a professional-services preset.
Inside
Exports SQLite · Postgres
Use
For AI labs, agent and RL environment builders, eval teams, and accounting and ERP software teams.
A 10-task sample is available on request.
Request the 10-task sampleProduct 03
Custom data and AI work, scoped to a request. Public-record data feeds are compiled from published sources and labeled as public records.
State transportation department asphalt contract records: lettings, bids, asphalt line items, and binder price indexes. The sample was compiled from public bid documents. Rows link back to the source document.
This sample is real public-record data. It is not synthetic.
02/Process
01
Share the product, the fields or the tasks, the volume, the format, and the use case.
02
A synthetic dataset or environment is generated to that specification. A public-record feed is collected from the published source and structured.
03
The schema, the labels, the reconciliation, or the source check is recorded, along with the methods and the limitations.
04
Synthetic files ship marked as synthetic. Listed datasets are sold on Datarade. Environments and custom work are sent on request.
These points apply to synthetic datasets and to Company in a box.
Public-record feeds are compiled from published sources. They are not synthetic, and the delivery says so. The DOT asphalt sample is one of those feeds.
Datarade's synthetic data policy asks for the same three things: disclose synthetic data in the title and description, label it clearly, and document the generation methods and algorithms. Publishing policies, section 4.4.
03/Custom
Do not include real personal data in this request.