Learning after launch on the AI Learning track. Production data drifts. Retraining on a schedule or on a trigger is part of learning, not a failure. Blindly training on live labels that are delayed or biased will chase the wrong target.
This lesson assumes you already worked through Fine-tuning.
The idea in practice
Define a drift check and a retrain cadence. Hold out a fresh week as a test. Do not auto-deploy a retrain that loses to the current model.
A concrete check
goal = {
'track': 'AI Learning',
'lesson': 'Learning after launch',
}
checks = [
'input available at decision time',
'score matches the real decision',
'failure case written down',
]
print(goal['lesson'])
for item in checks:
print('-', item)
Run the sketch locally if you have Python. The printout is a reminder of the checks, not a trained model. Replace the strings with the real inputs from your own example before you treat it as a design.
What usually goes wrong
Nightly training on clicks when clicks are not the outcome you wanted. When this happens, stop adding parameters or tools. Fix the check, the data, or the permission, then run the same example again.
What to write down
- The input you are allowed to use at decision time.
- The output and the score or pass rule.
- One failure you will test on purpose.
- What you will not claim the system can do.
Practice
Name one signal that should trigger a retrain review, not an automatic deploy.
Self-check
- Say Learning after launch in one sentence that mentions an input and an output.
- Name the failure mode in this lesson and the check that would catch it.
Done when: you can explain this lesson without the page open, and you have a written failure case.