HardSystem designSystem design30 min to answer

How would you serve a model under a strict p99 latency SLO?

1Give yourself 30 minutes

2Answer out loud, not in your head

3Then compare with the answer below

Stuck? Show a way to structure it+

01Clarify the objective, user, constraints, and acceptable failure modes.
02Structure the response around profiling, capacity, batching trade-off, degradation.
03Compare credible alternatives and state the decision criteria.
04Finish with validation, monitoring, and what would change the decision.

Reference answer

strong answer

I would first define the outcome and acceptable error boundary. I would explicitly cover profiling, capacity, batching trade-off, degradation. I would select an approach only after comparing quality, latency, cost, safety, and operational complexity, then validate it with representative cases and production monitoring.

profiling
capacity
batching trade-off
degradation

Must be adapted to the interviewer’s scenario
Would benefit from concrete numbers and personal examples

alternative answer

A second valid approach is to begin with the highest-risk assumption and test it cheaply. I would explicitly cover profiling, capacity, batching trade-off, degradation. I would select an approach only after comparing quality, latency, cost, safety, and operational complexity, then validate it with representative cases and production monitoring.

profiling
capacity
batching trade-off
degradation

Must be adapted to the interviewer’s scenario
Would benefit from concrete numbers and personal examples

Grade yourself

profiling

weight 30%

StrongExplains profiling concretely and ties it to a decision or measurable outcome.

WeakMentions profiling without implications, trade-offs, or evidence.

capacity

weight 25%

StrongExplains capacity concretely and ties it to a decision or measurable outcome.

WeakMentions capacity without implications, trade-offs, or evidence.

batching trade-off

weight 25%

StrongExplains batching trade-off concretely and ties it to a decision or measurable outcome.

WeakMentions batching trade-off without implications, trade-offs, or evidence.

degradation

weight 20%

StrongExplains degradation concretely and ties it to a decision or measurable outcome.

WeakMentions degradation without implications, trade-offs, or evidence.

Then expect these follow-ups

What assumption is most fragile?
Tests: depth
What alternative did you reject and why?
Tests: adaptability
How would you validate this in production?
Tests: adaptability

Hiring now

Machine learning engineer roles are open right now

landed applies to roles like these for you, on each company’s own site, with answers written for the role.

Get 5x interviews10 free auto-applies. No card.

AI/ML Engineer
Kaizen Labs · New York, NY
Senior ML Engineer, Perception
Clera · San Francisco
Machine Learning Engineer
Endava · Córdoba, Cordoba, Argentina

Or practice this question first

Next question

Design a batching service for model inference.