I interviewed at Nvidia in Yokneam, Israel.

The interview itself was fairly difficult and contained non-standard or trivial questions, including one that seemed more like a riddle than anything else. The interviewers were either disinterested or highly critical.

One interviewer was quiet almost the entire time. The other was very judgmental and dismissive, wanting to control the pace. When a question was answered correctly, he would not openly acknowledge it and would move on instead. He was not above throwing snide and unkind remarks and constant criticisms over minutiae unrelated to the content of what I was doing (e.g., he went on a random tangent about compilers for no apparent reason). He was also prone to prolonged silences.

Midway through the interview, I was stumbling through implementation details despite having done everything correctly. This was purely because I received no positive feedback at any point. I did not know that my solution was optimal or even on the right track until after I finished and checked things on my way back home.

I felt that the interview was not only difficult, but the interviewers made it harder than it needed to be. By the time I left, I felt crushed and my confidence was shaken, which is something that had not happened to me before.

Implement in C a routine that works such that you can, via a function, provide it with a time T and a void* callback. In T time from now, the callback will be automatically called. Specifically, optimize the solution for the first 256 units of time, since most scheduling will be done for them. To somewhat rephrase things, the system simulates several timers at once using its single internal clock, optimized around a specific time. The worst-case scenario can still be O(log(n)), but the most common/average case of times between 0 and 256 needs to be O(1).

The second question was more like a riddle:

You have 1000 wine bottles. Exactly one bottle is poisoned.

A rat that drinks the poisoned wine will die within 1 hour. (Assume rats can each sip any number of samples once at the start.)

Determine the minimum number of rats needed to identify which single bottle is poisoned in 1 hour.