How do I practice coding and algorithm questions?

Use PracHub's coding console to write, test, and debug your solutions in Python or JavaScript. View hints, test against sample inputs, and compare with official solutions.

What difficulty level is this coding question?

This is a Medium difficulty Coding & Algorithms question, commonly asked during Onsite rounds at Meta.

What role is this question designed for?

This question is commonly asked for Data Scientist candidates at Meta during technical interviews.

Compute 95th-percentile call concurrency

Quick Overview

This question evaluates a candidate's ability to design scalable algorithms for computing the 95th-percentile of per-minute concurrent calls over a 24-hour window, testing competencies in streaming/event aggregation, de-duplication, percentile computation, and managing time and memory constraints.

Company: Meta

Role: Data Scientist

Category: Coding & Algorithms

Difficulty: Medium

Interview Round: Onsite

Given N call sessions as half-open intervals [start, end) in UNIX seconds, design an algorithm to compute the 95th percentile of per-minute concurrent calls over the last 24 hours ending at 2025-09-01 00:00:00 UTC (i.e., minutes from 2025-08-31 00:00 to 2025-09-01 00:00). Requirements: (1) Input is an unsorted stream that may not fit in memory; (2) The same call_id may appear multiple times; treat duplicate intervals as one after de-duplication by (start,end); (3) A call active at timestamp t contributes to the minute floor(t/60); if an interval ends exactly at t, it does NOT contribute to that minute; zero-length intervals contribute nothing; (4) Return the 95th percentile (nearest-rank) of the 1440 per-minute concurrency counts; (5) Target time complexity O(n log n + M) where M=1440, and sublinear extra memory beyond storing the sweep-line events; (6) Describe your data structures (e.g., external sort + sweep or streaming bucketing), tie-breaking, and how you would extend it to compute the same metric per country if each interval carries a country code. Provide clear pseudocode and justify correctness.

Quick Answer: This question evaluates a candidate's ability to design scalable algorithms for computing the 95th-percentile of per-minute concurrent calls over a 24-hour window, testing competencies in streaming/event aggregation, de-duplication, percentile computation, and managing time and memory constraints.

Given half-open call intervals and a minute-aligned window, de-duplicate identical intervals, count how many calls overlap each minute bucket, and return the nearest-rank 95th percentile of those per-minute counts.

Constraints

Intervals are [start, end) UNIX-second pairs.
windowStart and windowEnd define the half-open analysis window.
Identical (start, end) intervals are counted once.
A minute bucket is counted if an interval overlaps any timestamp in that minute.

Examples

Input: ([[0,60],[0,120],[60,180]], 0, 180)

Expected Output: 2

Explanation: The three minute buckets have counts 2, 2, and 1.

Input: ([[0,60],[0,60],[30,60],[60,120]], 0, 120)

Expected Output: 2

Explanation: Duplicate intervals are ignored and end boundaries are half-open.

Input: ([[600,660]], 0, 180)

Expected Output: 0

Explanation: Intervals outside the window contribute nothing.

Input: ([[0,600],[120,300],[240,480],[540,660]], 0, 600)

Expected Output: 3

Explanation: Overlapping intervals are accumulated with a sweep line.

Hints

Clip intervals to the requested window before bucketing.
Use a difference array over minute buckets, then sort the counts for nearest-rank percentile.

Quick Overview