The setting
is a financial technology company that lets people buy now and pay later at checkout, and gives merchants payment tools. Its let shoppers split a purchase into scheduled payments.
Checkout's rising compute bill collides with headroom reserved for holiday bursts.
Capacity reviewers want spending tied to demand, while checkout engineers want short peaks protected from decisions based on averages.
Who you’d be doing this for
“I picked my payment option, then waited at checkout wondering whether clicking again would start another order.”
Fatoumata Fadahunsi · Retail shopper · a character in this brief
Uses an interest-free plan for purchases during busy shopping periods and depends on checkout remaining responsive.
What is at stake
A checkout service's compute bill rose 29% while requests grew 9%. You decide whether AI can draft a capacity change from observed peaks or must ask engineers to establish the burst allowance first.
Why it isn’t already fixed
Every obvious fix costs something else. That’s the part you’d have to decide.
- billable savings vs spare capacity for bursts
- service attribution vs shared node economics
- recent observations vs holiday demand
- context-sensitive proposals vs reproducible rules
Why Affirm
Shared Amazon EKS capacity supports checkout while Karpenter determines whether smaller workload requests actually release paid capacity.
Written with these in mind
Not your kind of problem? 128 more at Affirm, or browse every organization.
Where the organization context comes fromThe postings this brief was written from, and the reporting read before it.
The organization context comes from public information. The situation and characters are written for the brief.
Job postings it was written from
Public descriptions of the work at Affirm. They are listed as sources, not as openings. The lines under a posting are its own words, where it meets this brief.
- Senior Software Engineer, Backend (Workload Safety) (opens in a new tab)
- Affirm runs its workloads on more than 85 Amazon EKS clusters, and the number is growing.
- Determine the safe operating limits of each cluster (pod count, node count, API server request volume), keep clusters within them, and develop control plane scaling patterns for bursty workloads such as scheduled batch jobs and large deployments
- Senior Staff Software Engineer, Backend (Compute Platform) (opens in a new tab)
- As a member of the Compute Platform organization, you will collaborate with other teams - including Product, Infrastructure, Traffic, Cost & Capacity Engineering to
- Define and enforce EKS workload safety standards covering resource configuration, workload isolation and blast radius, disruption tolerance (PDBs), Karpenter consolidation strategies, etc
- Staff Software Engineer (Continuous Integration) (opens in a new tab)
This is the setup. The work is inside.
Running it puts you in the room: the full situation and its constraints, stakeholders who push back in their own words, and the decisions that are yours to make. What you produce becomes a Day One Plan — work you can show someone instead of describing.
Where similar work appears
Other live briefs connected to this one, and what connects them.
Similar problems at other organizations
Like this one, each of these turns on deciding on evidence that's still thin.
One inference-serving service is raising cloud spend while customer contact volume stays steady.
You must decide whether AI recommends capacity changes or only prepares evidence for each review.
Package-select conversion is falling, but the initial AI analysis used an invalid shopper cohort.
You must decide who confirms each cohort definition and comparison the agent builds from Snowflake and Amplitude before a flag reaches PM review in Slack.
Workflow failures take 45 minutes to turn into an engineer-approved investigation summary.
You have to decide where the AI stops reading CloudWatch and Datadog evidence and drafting failure paths, and where an engineer takes over the root-cause call and any change to jobs or retries.
The same setting, from another role
Each is set around Affirm’s Interest-free payment plans (Pay in 4, Pay in 2, Pay in 30), written for a different role. They’re separate briefs, not the same scenario.
A referral test must win 2,000 first-time plan selections while first scheduled-payment completion stays at 95% or higher.
You have to weigh low-cost acquisition against the clarity shoppers need before they commit.
A major home merchant may move its support toward a competitor as shoppers question fixed payment schedules.
You must weigh an immediate checkout lift against the trust signal the campaign leaves behind.
Campaign-referred shoppers select interest-free plans 11% of the time after reaching checkout.
You must weigh a stronger acquisition message against an expectation the fixed schedule cannot meet.