DevOps Engineer Interview Experience at Sprinklr – 2026

The recruiter screen covers current infrastructure ownership scope and confirms familiarity with observability tooling generally, since the team relies on a specific combination of PagerDuty for alerting and Graylog for log aggregation rather than a single all-in-one vendor.

Candidates are asked to design an alerting strategy for a fleet of services spanning several distinct product lines (social listening, marketing automation, and contact-center tooling), where a single unified alert dashboard has historically caused alert fatigue since one team's noisy-but-harmless alert pattern drowns out another team's genuine incidents. Per-product-line alert routing with a shared incident-severity taxonomy — so on-call engineers can still triage cross-product incidents consistently — is a strong direction, along with a weekly alert-noise review to actively prune rules that fire often but rarely correlate with a real incident.

This scripting problem asks candidates to parse a log file with mixed-format timestamps (some ISO 8601, some Unix epoch) and compute the P95 latency of requests grouped by endpoint over a rolling 5-minute window. Normalizing all timestamps to Unix epoch first, then using a sliding window per endpoint with a sorted structure to efficiently recompute the percentile as the window advances rather than re-sorting on every new event, is the expected approach.

The hiring manager asks about a time an on-call tradeoff had to be made between fixing a root cause immediately versus applying a quick mitigation and following up later, and specifically what factored into that decision. Prioritizing a fast mitigation during business hours when customer impact is ongoing, then scheduling proper root-cause work afterward, is a strong answer — along with communicating that tradeoff transparently to stakeholders rather than silently deferring the real fix.