Machine Learning System Design Interview, issue 9, Nov 28, 2025

The Catastrophic Forgetting Trap

AI Research Engineer interview at OpenAI, and the interviewer asks:

We deleted our original training dataset for GDPR compliance. We need to teach the live model a new class of data today. How do you do it?

Why fine-tuning on new data without old samples silently erases your model - and how EWC saves the day.

The full answer, with the mechanism and the arithmetic, is for paid subscribers on Substack.

Read it on Substack

Get the next one

Free on Substack. Unsubscribe in one click.