What is shadow or dark launching a model?
PICTURE THIS: HOW TO EXPLAIN IT
Simple meaning
Shadow serving sends live features to the new model but does not affect the user
WHY — Canary Deploy instead of guessing?
Why interviewers care about Canary Deploy:
people who only read docs from people who shipped.
and tied to MLOps work.
Name the idea, why it exists, then one short example.
End with when you use it and one common pitfall.
STEPS — What happens step by step?
Before you speak the answer, walk the interviewer through these steps:
- 1Shadow serving sends live
features to the new model but does not affect the user
- 2Why it exists
you only log the comparison.
- 3It validates latency and
prediction diffs with zero product risk.
- 4You still must protect
PII in those logs and not double-charge downstream systems.
- 5Common mistake
What juniors usually get wrong.
- 6Close
When you pick this over the alternative.
EXAMPLE — See it in action
Here's a short line you can speak, broken into clear beats:
Note: Adapt this scaffold to your own project — keep it under 60–90 seconds.
Key takeaway
Shadow serving sends live features to the new model but does not affect the user you only log the comparison.