1
Build the capacity model
Compute replicas from peak traffic, measured capacity, and 30% headroom.
Required evidence: Passing capacity calculation and assumptions table.
0/80 minimum characters
Advanced mini-project·3-4 hours
ML Serving Readiness Project
Turn quality evidence and traffic assumptions into a capacity model, SLO, rollout, and rollback plan.
Scenario
A model endpoint must serve 120 requests per second at peak while meeting a p95 latency target. One warm replica sustains 35 requests per second at the target latency.
You will demonstrate
- Compute capacity with headroom.
- Define complete serving SLOs.
- Design canary and rollback evidence.
Project evidence
Show the work, not a checked box.
Each response is stored in this browser as you type. Include metrics, test output, or a decision rationale wherever the deliverable asks for it.
2
Specify the SLO
Define latency percentile, error rate, availability, quality, and observation window.
Required evidence: A measurable SLO contract.
0/80 minimum characters
3
Design the load test
Include warm, cold, burst, overload, and dependency-failure cases.
Required evidence: A reproducible load profile and acceptance criteria.
0/80 minimum characters
4
Plan canary and rollback
Choose canary size, guardrails, rollback trigger, and fallback.
Required evidence: A staged release runbook.
0/80 minimum characters
Runnable Python lab
Calculate replica capacity with headroom
Compute the minimum whole number of replicas for peak traffic and a utilization headroom policy.
Project defense
Why is average latency insufficient for an online inference SLO?
Rubric self-review
Rate the evidence, not your effort: 0 missing, 1 weak, 2 adequate, 3 strong. All criteria must be reviewed, but a low honest score does not get hidden.
Capacity includes explicit headroom.
Tail latency and queueing are measured.
Quality remains a release dimension.
Rollback is automatic for defined critical failures.
Useful references
Project completion gate
Completion is controlled by stored evidence, deterministic tests, a decision defense, rubric review, and the artifact when required.
Evidence pendingCode pendingDefense pendingRubric pending
Optional cloud portfolio
Submit evidence across devices.
An account is required. Submit only when the local completion gate passes. AI review is advisory and separate from deterministic completion.