Brex DevOps / SRE Interview Questions

1.Tell me about an incident or outage you inherited mid-flight — one that wasn't your team's system but you ended up owning the resolution anyway. How did you step in, and what did the outcome look like?

2.Walk me through a written proposal or design doc you authored that changed how your team operates — maybe an RFC, runbook overhaul, or architecture proposal. What was in it, and how did it land?

3.Describe a time you pushed back on a quick fix — a config change, a hotfix deploy, or a workaround — because you believed it would create a bigger mess down the road. What was the pressure you were pushing against, and how did it resolve?

4.Tell me about a time you surfaced a production reliability risk — a fragile dependency, a capacity ceiling, a toil-heavy process — before it became an incident. Who did you tell, how did you communicate it, and what happened?

5.Walk me through a time you took over a system or service that was poorly documented, fragile, or had a lot of tribal knowledge — and you had to stabilize it without a formal mandate to do so. What did you actually do, and where did you start?

6.Tell me about a design doc or RFC you wrote where you proposed a significant change to how your team or org handles deployments, incident response, or infrastructure — and the proposal was controversial. How did you build the case in writing, and how did you handle the disagreement?

7.Our Kubernetes clusters run a mix of latency-sensitive gRPC services and batch jobs. How would you set up resource requests, limits, and priority classes to make sure a noisy batch job doesn't starve a card authorization service?

8.You're looking at a Kafka consumer group that processes expense events for Brex Spend Management and the lag is growing. It's been fine for months. How do you approach diagnosing and resolving it?

9.We're running a PostgreSQL database for a service that handles Brex Business Account ledger writes. Writes are starting to slow down under load and you're seeing lock wait timeouts. Walk me through how you'd investigate and what solutions you'd consider.

10.You need to design a zero-downtime deployment pipeline for a gRPC service that handles card authorization requests — it needs to process thousands of transactions per second with no dropped connections during a deploy. How would you build this?

11.We want to enforce that every microservice at Brex emits a standard set of spans and metrics — service name, trace ID, error codes, and latency histograms — but we have 200+ services across Kotlin, Elixir, and TypeScript. How would you actually enforce this at scale without making it a manual review burden?

12.Brex's card authorization service needs to remain available even if our primary AWS region goes down. Walk me through how you'd architect multi-region active-active or active-passive failover for a service that reads and writes to PostgreSQL on every transaction.

13.We need a health-check and readiness-probe system for Brex's microservices on Kubernetes. How would you design it so that a card authorization service pod only receives traffic once it's genuinely ready to handle requests — not just that the process started?

14.Brex needs a centralized secrets management system — API keys, database credentials, third-party payment network tokens — that our 200+ microservices can access at runtime. How would you design it?

15.Walk me through how you'd design an on-call alerting system for Brex's spend management platform — one that pages the right team for real incidents without burning people out on noise. What does the full pipeline look like?

16.Design a rate-limiting layer that sits in front of Brex's external-facing APIs — the ones Brex Card program managers and Brex Empower enterprise customers call to manage spend policies and card controls. It needs to be enforced at the edge with sub-millisecond overhead. How do you build it?

17.Design a cost attribution and showback system for Brex's AWS infrastructure — one that lets engineering teams see what their services actually cost, tied back to specific products like Brex Card, Brex Business Account, and Brex Empower. How would you build this end to end?

18.Design a global configuration delivery system for Brex — one that pushes spend policy rules, card controls, and feature flags to services handling Brex Card authorization and Brex Empower enterprise policies, with sub-second propagation and strong consistency guarantees across regions. Walk me through the architecture.

19.Tell me about a time you had to ramp up on an unfamiliar service or codebase quickly. How did you orient yourself, and what did you produce to help the next person do the same?

20.Describe a time when your team's on-call rotation was burning people out. What did you do about it, and how did you know if it worked?

21.Tell me about a time you had to get a team that wasn't yours to prioritize a reliability or infrastructure change. How did you build the case and what happened?

22.Walk me through a time you made a call during an incident that turned out to be wrong — maybe a rollback that caused more problems, or a fix that masked the real issue. How did you handle the aftermath?

23.Tell me about the most complex migration you've driven — moving a service to new infrastructure, changing a data store, or shifting traffic to a new system. How did you plan it, and what would you do differently now?

24.You're three months into a role and you've noticed that your team's deployment process has a serious gap — no automated rollback, manual approvals that create bottlenecks, unclear ownership between SRE and product engineering. Nobody has asked you to fix it. What do you do?

25.Estimate how many Brex Card authorization requests we'd process on Cyber Monday compared to a normal Tuesday. Walk me through how you'd size infrastructure for that spike.

26.Your team gets paged because error rates on the Brex Business Account money movement API spiked to 8% — up from under 0.1%. It's been 12 minutes and nothing obvious shows in the deployment history. How do you structure the next 30 minutes?

27.Brex runs 200+ microservices across Kotlin, Elixir, and TypeScript. Estimate what percentage of your total AWS bill is likely attributable to idle or underutilized compute — and walk me through how you'd validate that estimate.

28.Brex's expense ingestion pipeline needs to handle a 10x traffic spike during end-of-quarter — finance teams at enterprise customers are submitting and reconciling expenses in bulk. How would you design autoscaling for this pipeline so we handle the spike without over-provisioning year-round?

29.We're considering adding a synchronous compliance check — calling an external sanctions screening API — inline to the Brex Card authorization path. Each call adds roughly 80ms of latency. Estimate the full downstream impact on our authorization SLA and infrastructure cost, and tell me whether you'd recommend doing it synchronously or not.

30.Brex's SRE team is responsible for 200+ microservices but only has a headcount of, say, 15 engineers. Walk me through how you'd decide which services the SRE team should own deeply versus which ones you'd push back to product engineering teams — and how you'd enforce that boundary without creating gaps.