How would you diagnose a sudden p99 spike on a model API?
PICTURE THIS: HOW TO EXPLAIN IT
Simple meaning
I would check deploy correlation, GC or Python GIL stalls, feature store latency, GC of large batches, and downstream timeouts.
WHY — Latency and Throughput instead of guessing?
Why interviewers care about Latency and Throughput:
question about Latency and Throughput.
trade-offs, and what you would actually do on a MLOps project - not buzzwords.
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:
- 1I would check deploy
correlation, GC or Python GIL stalls, feature store latency, GC of large batches, and downstream timeouts.
- 2Trace IDs from the
API into Redis or the warehouse matter.
- 3Scaling CPU replicas will
not help if the online store is the bottleneck.
- 4Give an example
One tiny concrete case you can say aloud.
- 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
I would check deploy correlation, GC or Python GIL stalls, feature store latency, GC of large batches, and downstream timeouts. Trace IDs from the API into Redis or the warehouse matter.