DevOps Engineer Interview Experience at Cult.fit (Cure.fit) – 2026

An online coding round on HackerEarth with two problems - a dynamic programming question on computing the minimum cost to reach the end of an array with variable step costs, and a scripting-style problem to parse a stream of class-booking events and compute, per class, the maximum concurrent number of confirmed attendees at any point in time. The expected approach uses a sweep-line technique over booking-start and cancellation events sorted by timestamp.

Candidates are asked to design capacity management for a popular live-streamed workout class scheduled to start at a fixed time, where a large number of users attempt to join within the same narrow window right before the class starts. A strong approach pre-warms compute capacity ahead of the scheduled start time based on the number of confirmed bookings, rather than relying purely on reactive autoscaling, since reactive scaling triggered only after the traffic spike begins would be too slow for a spike this sharp and predictable. Interviewers push on what happens if a class unexpectedly goes viral beyond its normal booking numbers; a graceful-degradation fallback (dropping to lower video quality automatically under sustained overload) is preferable to an outright service failure.

A genuinely conversational round built around a take-home infrastructure design assignment submitted earlier in the process (designing a monitoring and alerting setup for the class-booking service). Rather than a rigid Q&A, the interviewer engages directly with specific tradeoffs in the candidate's design, particularly around alert threshold tuning for a system with naturally bursty, event-driven traffic, and shares their own perspective on how the real production system handles that same tension - it tends to feel more like a peer design review than an evaluation.