Intervals & Scheduling

Interval problems are a light cluster on NVIDIA's algorithm track, and when one lands the algorithm is rarely the hard part. Two problems cover almost all of it: merging overlapping intervals into the fewest ranges, and counting the largest number of jobs running at the same instant. Both are mediums, and NVIDIA's medium bar reads as approachable next to Google or Meta. The difficulty lives a layer down, in the C++ and performance follow-ups the interviewer stacks on once the ranges are sorted.

That layer is where the round is decided. Expect to defend an O(n) sweep by counting how many times memory gets touched. NVIDIA interviewers often ask you to build the container by hand, so "keep a min-heap of end times" can turn into "write that heap, no STL." The concurrency count is native ground here too. How many meetings overlap at once has the same shape as how many jobs a fixed pool of GPUs can run at once. An interviewer reads a clean answer to it quickly.