The Training-Serving Skew Trap
When the data at serve time stops matching training data.
What Skew Means
Training-serving skew is when the data your model sees in production no longer matches the data it learned from in training. 📉
Why It Hurts
Your model assumes serving data looks like training data. When that assumption breaks, predictions quietly get worse even though nothing crashes. This is skew.
All lessons in this course
- The Training-Serving Skew Trap
- Silent Failures: No Crash, Wrong Answers
- When the World Changes Under Your Model
- The Reproducibility Problem