“I’d give them a real architectural decision that already has ambiguity and trade-offs, like whether to split a service or keep it monolithic, and watch if they ask about team size, deployment cadence, and blast radius before drawing a line. Ownership shows up when someone names the thing they will be accountable for, not just the cleanest diagram.”

“I look for someone who can defend a decision under pressure, not just draw boxes on a whiteboard. At Fidelity, the architects I trust most are the ones who have carried a bad production call and can explain why the trade-off was worth it.”

“I look for the engineer who, when I describe a messy integration or a target architecture problem, starts drawing the boxes and asking about the lifecycle before they ask about the stack. At Valassis, the person who could own the IMO engine wasn’t the one with the cleanest code; it was the one who could hold the whole media-optimization flow in their head and argue about the seams.”

“I’d give the candidate a real architectural problem we’re facing and ask them to sketch the system, then push on tradeoffs, failure modes, and what they’d defer. At Fidelity, the best hires were the ones who could name what they didn’t know before they started designing.”

“I look for the candidate who, when I describe the ugliest current system decision, starts sketching how to unwind it rather than how to replace it wholesale. At Tala and Ring, the architects I trusted were the ones who inherited a mess and made it incrementally better before they ever got to draw the clean diagram.”

“I look for the candidate who can explain the worst decision they made on a system that lived for years, not the one who draws the cleanest diagram in the interview. Ownership shows up when someone has paid for a wrong call and still has to maintain it.”

“I look for the candidate who, when I describe a messy constraint, starts sketching trade-offs out loud instead of asking for a spec. At Apozy, the senior hires who stuck were the ones who would whiteboard three bad options and pick the least bad one before I finished my coffee.”

“I watch for the candidate who asks what we are not building, not just what we are. At SentinelOne, the best architecture owners I worked with started by naming the constraints we would regret ignoring in six months.”

“I look for the candidate who can explain why they chose one data structure over another in a system they actually built, not the one with the cleanest diagram. At Orama, the people who owned architecture were the ones who could argue about trade-offs under load and still listen to users.”