Meta Infrastructure Engineer Interview Questions

Implement a min-heap of size k. Iterate through the array: add elements if the heap has fewer than k items, otherwise only add an element if it's larger than the heap's minimum (root). After processing all elements, the heap's root is the kth largest. The key insight is that you only need to maintain k candidates, not sort the entire array. This achieves O(n log k) time and O(k) space. QuickSelect offers O(n) average time, but its worst case is O(n²), making the heap approach more reliable for interviews.

Use a single-pass greedy approach: track unmatched opening parentheses with a counter. For each '(', increment it; for each ')', decrement if counter > 0 (matched), otherwise count as a removal. At the end, any remaining unmatched '(' must also be removed. The key insight is that we need only counters, not storage—every '(' without a matching ')' and every ')' without a preceding '(' must be removed.

Build a cumulative weight array where each index stores the sum of weights up to that point. To select: generate a random number in [0, total_weight), then binary search the cumulative array to find the first index where cumulative_weight ≥ random_value. This transforms O(n) linear scanning into O(log n) queries after O(n) preprocessing. The key insight is that cumulative weights map the weight distribution to a continuous range we can search efficiently.

Use three pointers (one per array) to track your position in each. At each step, compare the current elements from all three arrays and append the smallest to the result, advancing that array's pointer. When an array is exhausted, continue merging the remaining two. This greedy approach processes each element exactly once, making it optimal.