Airbnb Intern OA, Diplomacy Game: 19 Experiences (Part 4 of 7)

Shared by janebeingjane ·
Company
Airbnb
Level
Internship
Stages
OA

Airbnb SWE intern OA reports from 2016 to 2019: one task per OA, mostly the Diplomacy battle game in fall 2018 and host-crowding pagination in fall 2017.

At a glance

  • Process: One coding task per OA, under a label that matched the task: "OA2" or "code challenge 2" in fall 2017 (H58, H60, H61), "OA4" in fall 2018 (H48, H50, H51, H52, H56) and "2019.4" in September 2019 (H45–H47). Where a window was given it was a few days: four days in H48, H51 and H53. H52 and H53 had one hour.
  • Asked in these reports: The Diplomacy simulation led fall 2018 with 5 reports (H56 by its OA4 label), and host-crowding pagination led fall 2017 with 5. The three 2019 reports each got a different task: the decimal task, module costs and the XML validator.
  • Difficulty: H52 found the Diplomacy task not hard, with time to spare. Others had it tighter: H53 rated its OA medium to hard, H59 called the time limit tight, and H55 ran out of time.
  • Practise first: Resolve Battles and Paginate Search Results by Unique Host.

Coverage: 19 software engineer intern candidates who reported only the OA. Three are from September 2019 (H45–H47), ten from fall 2018 (H48–H57), five from fall 2017 (H58–H62) and one from 2016 (H63). Of the 11 who gave a degree, 8 were master's and 3 bachelor's students. Eight gave a United States location, two of them a city (Seattle, H54; Durham, North Carolina, H63). Three passed the OA (H46, H52, H56), H61 missed its invite email, eight did not pass or got no offer (H45, H49, H53, H55, H57, H59, H62, H63), H54 was still waiting, and the rest did not say.

In this series

What came up

Each fall had its own task, and the OA label showed which. Counts are out of the 19 reports in this part; four reports named no task (H53, H55, H57, H63).

Task Reports Last seen
Resolve Battles (the Diplomacy game; H56 by its OA4 label) 5 of 19 Dec 2018
Paginate Search Results by Unique Host (host crowding) 5 of 19 Nov 2017
Fraction to Recurring Decimal (LeetCode 166) 1 of 19 Sep 2019 (report date)
Module Rebuild Costs 1 of 19 Sep 2019
Simplified XML Validator 1 of 19 Sep 2019 (report date)
A parse function, not described 1 of 19 Dec 2018
A game implementation, not described 1 of 19 Oct 2018

History (before 2023): 19 OA-only reports

H45: Division, one random question

Report date: Sep 2019 · Role: SWE intern (master's student) · Result: Did not pass

  1. Referral. The screen read "Airbnb Coding Challenge - New 2019.4" and served one randomly chosen question. This candidate's was called Division, which is Fraction to Recurring Decimal (LeetCode 166, described under H12 in Part 2).
  2. This version said both inputs are positive integers. A finite answer comes back as is (6 / 4 gives "1.5"), and a repeating part goes in parentheses (7 / 3 gives "2.(3)"). The division operator was allowed. More examples: 2 / 4 → "0.5", 6 / 3 → "2", 1 / 3 → "0.(3)" and 229 / 990 → "0.2(31)".
  3. The shared screen showed 15 minutes left and the question not yet attempted.

H46: Module costs as a graph count, moved on

Interview date: Sep 2019 · Role: SWE intern (master's student) · Result: Passed the OA

  1. Referred on 3 September. The OA arrived on 17 September and was done on 20 September; a move-on notice followed. The candidate called it version 2019.4.
  2. Their framing of Module Rebuild Costs (worked example under H21 in Part 2): in a directed graph, changing node x affects every node that depends on it. Output that count for every node.
  3. The approach was DFS with a cache that holds, for each node, a set of nodes (itself and everything above it), not a number. The answer is the size of the union, so a node reached by two paths counts once.

H47: XML validator, and others got different problems

Report date: Sep 2019 · Role: SWE intern (bachelor's student)

  1. Referred about two weeks earlier. The "New 2019.4" OA gave the Simplified XML Validator, with the rules under H31 in Part 3.
  2. A stack did it, returning a different string for each kind of error. The candidate found it direct, and noticed other people getting different problems.

H48: The army task, then no reply

Interview date: Dec 2018 · Role: SWE intern (master's student)

  1. Met Airbnb at a campus career fair. The OA was "OA4", the army one: Resolve Battles, the Diplomacy game described under H52.
  2. Four days to finish: it arrived on 18 December and went in on the evening of 21 December.
  3. No news followed, even after two follow-ups to the recruiter. The candidate had heard only two tests are visible on screen.

H49: A parse function, all tests passed

Interview date: Dec 2018 · Role: Software Engineer (Internship) · Location: United States · Result: Rejected after the OA

  1. An OA link with several days to finish. It included a parse function, slightly different from versions the candidate had seen before; input and output not described.
  2. Every test passed, and a rejection came anyway.

H50: Diplomacy, input and output as string lists

Interview date: Nov 2018 · Role: SWE intern (master's student)

  1. Met a recruiter at a campus career fair on 5 September and applied that day. "OA4" came only on 8 November.
  2. Still the army game, Resolve Battles (rules under H52). What this report adds: input and output are both lists of strings, and the output is sorted by army name. The candidate used a sorted map (a TreeMap) keyed by name.
  3. They had worried about support chains (A supports B, B supports C). No special case is needed: an attacked army cannot support anyone, and a few loops do the rest.
  4. Their big test: thirteen armies from different towns all move into one city, and a fourteenth supports one of them. That army, at strength 2, takes the city; the other twelve die and the supporter stays put.

H51: Diplomacy, and a flaky last test

Interview date: Nov 2018 · Role: SWE intern (master's student)

  1. Applied online on 23 October. "OA4" arrived on 8 November, and HR suggested finishing within four days.
  2. The task was the Diplomacy simulation, Resolve Battles (H52).
  3. The last test sat right at the time limit: the same code, unchanged, timed out once and passed on several reruns.

H52: Diplomacy rules, with the Munich example

Interview date: Nov 2018 · Role: SWE intern (master's student) · Result: Passed the OA

  1. Applied online on 22 October and took "OA4" on 5 November: one problem in one hour, which the candidate found plenty.
  2. The task, Resolve Battles: each input string is one army's order. A Munich Hold means army A stays in Munich. B Warsaw Support A means B, in Warsaw, adds 1 to A's strength, so A has 2. C Bohemia Move Munich means C attacks Munich from Bohemia with strength 1, so C loses and A stays. When armies meet in one place with equal strength, all of them die. An army that is attacked loses its support. The output lists each army's final state: A Munich, B Warsaw, C [dead].
  3. No tricks, only care. A single pass did not work out, so the candidate used two; nothing timed out.
  4. A phone-screen invite came the next day.

H53: One hour on HackerRank

Interview date: Oct 2018 · Role: Software Engineer (Internship) · Location: United States · Result: Did not pass the OA

  1. A HackerRank OA: four days to finish it, with one hour once started. The candidate rated it medium to hard and did not describe the task.

H54: A game to implement

Interview date: Oct 2018 · Role: Software Engineer (Internship) · Location: Seattle, Washington · Result: Pending when written

  1. A recruiter reached out, then sent a HackerRank OA. The task was a game implementation. The game and its rules were not described.

H55: A take-home that ran out of time

Interview date: Oct 2018 · Role: Software Engineer (Internship) · Location: United States · Result: Did not pass the OA

  1. A take-home coding challenge at about LeetCode medium level, not named. A few test cases passed before time ran out.

H56: "OA4", one case failed, still invited

Interview date: Oct 2018 · Role: SWE intern (master's student) · Result: Passed the OA

  1. Referred on 24 September and took "OA4" on 13 October. The report gave only the label. Other fall 2018 reports here say OA4 was the Diplomacy task (H48, H50, H51, H52), so it was most likely Resolve Battles.
  2. One test case failed, and a phone-interview follow-up came anyway, on the Monday after.

H57: Cut at the coding challenge

Interview date: Fall 2018 · Role: SWE intern · Location: United States · Result: Did not pass the OA

  1. In a post about a whole internship search, begun with no computer science background, Airbnb got one line: cut at the coding challenge. No format or task given.

H58: "OA2", host crowding

Report date: Nov 2017 · Role: SWE intern (bachelor's student)

  1. Applied online. "OA2" was one problem, host crowding: Paginate Search Results by Unique Host, explained under H61.

H59: Unique hosts on each page, tight on time

Interview date: Nov 2017 · Role: Software Engineer (Internship) · Location: United States · Result: No offer

  1. HackerRank, one question, any language, with a tight time limit.
  2. The task was paging entries with unique hosts: Paginate Search Results by Unique Host (H61).

H60: "Code challenge 2" was pagination

Report date: Oct 2017 · Role: SWE intern (master's student)

  1. Referral. "Code challenge 2" was the pagination task, Paginate Search Results by Unique Host (H61), and nothing more.

H61: Host crowding, and a missed email

Interview date: Oct 2017 · Role: SWE intern (bachelor's student)

  1. Applied online. "OA 2" came about two weeks before the report and was done a few days later: host crowding, Paginate Search Results by Unique Host.
  2. The rule as this candidate understood it: split the listings into pages of a fixed size. Each page avoids repeating a host id where it can, keeping the original order. When no new host is left for a page, fill it with the next listings in their original order. With host ids 1, 1, 1, 1, 2, 2, 2, 2 and pages of about four, their answer was 1, 2, 1, 1 and then 1, 2, 2, 2. For 1, 1, 1, 1, 5, 5, 5, 5, 5, 5 the first page was 1, 5, 1, 1.
  3. They used a hash set of hosts already on the page, a linked list and an iterator. In the OA they switched to a "better" spread (1, 2, 1, 2 twice), and one test failed. Afterwards they thought the plain order-keeping fill was wanted, and a reader agreed.
  4. A next-step email came the same day, after an engineer reviewed the code. It sat in spam for a week, the recruiter closed the application, and an appeal did not work.

H62: Splitting hosts into result pages

Interview date: Oct 2017 · Role: Software Engineer (Internship) · Location: United States · Result: Did not pass the OA

  1. A coding challenge on spreading hosts across search-result pages: the Paginate Search Results by Unique Host task (H61).

H63: A 2016 career-fair start

Interview date: Sep 2016 · Role: Software Engineer (Internship) · Location: Durham, North Carolina · Result: No offer

  1. Met Airbnb at a career fair. A recruiter phone call came first, then an online technical question.
  2. The candidate said the question needed a concept beyond first-year courses. Neither the concept nor the question was named.

Patterns across the reports

  1. The label gave the task away. Every fall 2018 report that named its "OA4" task named the Diplomacy game (H48, H50, H51, H52), and all three named fall 2017 "OA2" or "code challenge 2" tasks were host crowding (H58, H60, H61). September 2019 broke the pattern: under the same 2019.4 label, H45, H46 and H47 each got a different task, and H45's screen called it a random question.
  2. Test results did not decide the outcome. H49 passed every test and was rejected. H56 and H61 each failed one test and still got a next step.
  3. Reading too much into the rules. H50 worried over support chains that need no special case, and H61 swapped the plain fill order for a smarter spread and failed a test.

Keep practicing

In this series

Sources: Anywhere useful interview info can be found.

Next in this seriesAirbnb Intern OA and Onsite Rounds: 16 Experiences (Part 5 of 7)
0 total reactions. Your reactions: none.

Comments

0

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