RepostedBloomberg LP logo

Bloomberg LP · Phone screen

Rejected

Bloomberg SWE Phone Screens with Two Teams: 1 Experience

Software engineer · Full-timeMaster's degreeReported Nov 16, 2026
Shared byjanebeingjane

A 2026 Bloomberg experienced engineer's phone screens with two teams at once: coins, film recommendations and a sorted-set class; rejected by both.

At a glance

  • Process: phone screens with two teams in parallel, each 15 min resume + coding + 5 min questions (report 01)
  • Asked recently: fewest coins with the combination, film recommendations from similar viewers, a sorted add() class (report 01)
  • Difficulty: the candidate rated it average and the experience negative
  • Practise first: Fewest Coins Combination, Track New Values in Sorted Order

Coverage: One experienced engineer with a Master's degree, changing jobs, who applied online. The report covers technical phone screens with two teams at the same time. It gives October to December 2026 as the interview period and went up on November 16, 2026.

What came up

Bloomberg let the candidate interview with two teams at once, with independent results, and says the company requires finishing two coding questions. Team 1's screen had two; Team 2's had one, with a long follow-up. Each screen opened with about 15 minutes on the resume and ended with 5 minutes for questions.

Topic Reports Last seen
Coding: fewest coins with the chosen combination 1 of 1 Nov 2026 (report date)
Coding: recommend films from viewers whose history covers yours 1 of 1 Nov 2026 (report date)
Coding: track new values and return them in sorted order, then a memory follow-up 1 of 1 Nov 2026 (report date)

Questions we found

Algorithms / Data Structures

Question Reports Last seen
With unlimited coins of five denominations, return the combination that makes an amount with the fewest coins (practice) 1 Nov 2026 (report date)
Given each user's watched films and a new user's films, keep only users who watched everything the new user did, and recommend their other films by how many of them watched each (practice) 1 Nov 2026 (report date)
A class whose add(x) says whether x is new and returns all values so far in sorted order; follow-up on reducing memory for many calls (practice) 1 Nov 2026 (report date)

1 detailed candidate timeline, loops, and screens

01: Experienced SWE, phone screens with two teams, not moved forward

Interview period: 2026 Q4 · Report date: November 16, 2026 · Role: Software engineer · Degree listed: Master's · Channel: Online application · Candidate's rating: Negative, average difficulty · Result: Not moved forward with either team

Scheduling

The first team's screen was moved the night before with no formal email, and for the second team nobody joined at the scheduled time. Both were rebooked about a week later.

Team 1 (team lead)

  1. Coins. Five coin types (the candidate recalls values like 1, 5, 10, 15 and 20), each usable any number of times. For an amount n, output the combination with the fewest coins, for example three 1-coins for n = 3 and one 5-coin for n = 5. The candidate used unbounded-knapsack dynamic programming that also keeps the best combination for each amount. FastPrep's practice version is Fewest Coins Combination.
  2. Film recommendations. Users map to the films they watched. For a new user's list, keep every other user whose history contains all of the new user's films, then recommend the films those users watched that the new user hasn't, ordered by how many of them watched each. In the reported example the answer was [d, b, e]. The candidate checked containment, took the difference, counted and sorted. FastPrep's practice version is Recommend Films from Viewers Who Watched Everything You Did.

Both were finished and passed the tests, but the rejection came two days later with no reason given.

Team 2 (senior engineer)

The question was a class with add(x) that records an integer, says whether it was new or seen before, and returns all values so far in sorted order. The candidate noted that Java's TreeSet would do it and, in Python, kept a sorted list with binary search: O(log n) to find, but O(n) worst case to insert. The interviewer kept stressing that add is called very many times and questioned the memory use. The candidate proposed a bitmask with one bit per value in 32- or 64-bit words, noting it may waste space when values are sparse. The interviewer didn't say whether that was what they wanted, and there was no time to code it. FastPrep's practice version is Track New Values in Sorted Order.

Result

Neither team moved forward.

Keep practicing

Sources: Anywhere useful interview info can be found.

0 total reactions. Your reactions: none.
0 comments

Comments

0

No comments yet — be the first to share what you know.