Private data toolkit
Credit-risk data toolkit
Start with a local, read-only profile before building point-in-time datasets, labels, features, and monitoring tables.
Runs in this browser only
Profile a credit-risk CSV
Inspect types, missingness, repeated records, categories, dates, target balance, and simple numeric outlier flags. This page has no upload or save action: the file is parsed in browser memory.
Check the data grain
Confirm whether a row represents an application, obligor, facility, or account-period before joining labels or calculating rates.
Review time alignment
Keep feature availability dates, observation windows, default windows, and outcome maturity separate to reduce look-ahead leakage.
Treat flags as prompts
An IQR outlier, missing cell, or duplicate key needs investigation against the business definition; a simple profiler cannot decide whether a value is wrong.