Unit Test
Last Updated: July 29, 2026 | By Mihail Sebastian | AI Dictionary
An automated test that checks one small piece of code in isolation; in ML projects, the guard that keeps data pipelines and feature code correct.
What is a Unit Test?
A unit test is an automated check that one small piece of code, usually a single function, behaves as expected in isolation.
In machine learning projects, unit tests guard the code around the model: data loading, feature transformations, and pipeline steps. Bugs there rarely crash anything; they silently corrupt every prediction downstream.
How Unit Tests Work
A test calls a function with a known input and asserts the expected output:
def test_scale_to_unit_range():
assert scale([0, 5, 10]) == [0.0, 0.5, 1.0]
Tests run automatically on every code change, so a failure points straight at the unit that broke. The trained model itself resists unit testing, since its behavior is learned rather than written.
The deterministic code around it does not: encoders, tokenizers, date parsers, and threshold logic all have exact right answers worth pinning down in tests. That is where feature engineering bugs hide.
Example of a Unit Test
A fraud-detection team has a function that converts transaction amounts into a customer’s home currency before scoring. Its unit test feeds in known amounts and asserts the converted values to the cent.
Months later, a refactor accidentally inverts an exchange rate. No error appears anywhere; the numbers are simply wrong, and a model retrained on them would quietly lose accuracy.
Instead, the test suite fails within minutes of the change, naming the exact function. The developer fixes the inversion before the bad values ever reach model training, which is precisely the class of silent failure unit tests exist to catch.
Related AI terms: Pipeline · Feature Engineering · Data Quality · Version Control
Did you like the Unit Test gist?
Learn about 250+ need-to-know artificial intelligence terms in the AI Dictionary.
Mihail Sebastian — Writes about AI governance, regulation, and the technology behind them. Placeholder bio — replace with a real credential line. About