This problem, where a model fails on unseen (out-of-distribution) data, is best solved by designing a system that uses model uncertainty and user feedback to mine and label these "hard examples."
Had the beggining of the call with OS fundamentals on thread vs process, process synchronisation and the screening 2 Leetcode questions - 1 easy on Tree and 1 Difficult on backracing
Got a message from the recruiter on linkedIn that my profile has matched the JD and would like to setup a call.
A brief 15 min call with the recruiter. Then it was a 1 hour technical screening.
In the screening the developer asked 8-10 iOS hitters and then we dived into coding. Basic BFS and DFS question which we wrapped up in like 20-25 minutes. Then we just chatted and ended the interview.
Later the HR told me that LinkedIn s interview will be onsite for the super day and I d need to fly to Mountain View HQ for that.
I was eager for that and waited for the next step. After that the HRs just disappeared. I emailed and messaged her on LinkedIn but no luck.
Extremely disappointed.