The recruiter screen was first, a very standard and pleasant initial screening.

Next was a technical screen with a staff data scientist, who was also very pleasant. The discussion primarily focused on my experience as it related to the problems Okta is currently tackling. A straightforward coding question involving two pointers concluded the session (your first guess is likely correct).

This was a positive experience, offering ample opportunity to ask questions about Okta's expectations, which were answered cogently.

Following this was a virtual on-site with three technical rounds:

• The first round was a systems design interview with a principal architect. This was the best interview by far—a pertinent systems design question relevant to my experience and Okta's needs, plenty of time for discussing the role, and a very insightful interviewer.

• The next two rounds were not as positive.

• The fourth round was a coding challenge with an architect, involving sliding windows and data aggregations. The architect was not very friendly and allotted time for precisely one question on my part. The coding problem was at an easy level on LeetCode.

• The fifth round was a general ML experience interview with a staff data scientist. The first half covered topics like feature stores, model drift, and model deployment. The second half involved a coding challenge, which was the exact same problem as the initial coding screen, indicating a lack of coordination between interviewers. I provided the most optimal solution, and the interviewer seemed unprepared, spending 15-20 minutes asking me to consider less efficient solutions. This was followed by the interviewer openly googling questions for machine learning engineers during interviews and then posing them to me, such as "what is a significance test?" and "how would you set up an A/B test?", leaving time for precisely two questions on my part.

Overall, Okta does not seem to know how to interview ML engineers at this level.

Two out of the four interviewers were hyper-focused on coding problems at a level where basic competency should be assumed. They did very little to probe my experience or provide an opportunity for me to guide the interview, and generally seemed disinterested. When I asked about their work, the staff-level data scientist and the architect gave answers I would expect from mid-level professionals. I got the impression they were quite incompetent. Given this is the impression they are giving candidates, I would be wary of an offer, even if one were extended.

Two coding questions:

• One involving 2 pointers (yep, that one).
• One involving a sliding window plus minor data aggregations.