RepostedGoogle logo

Google · Phone screen → Onsite

In process

Google SWE and PhD Intern Coding: 14 Experiences (Part 3 of 3)

Software engineer · InternshipBachelor's degreeSep 2026
Shared by

Google Summer 2026 SWE intern technical rounds from Jan–Feb 2026: 14 reports, mostly two 45-minute coding rounds on graphs, DP, heaps and streams.

At a glance

  • Process: seven reports describe two coding rounds, often 45 minutes each and back to back (31–34, 36–38); one Hyderabad candidate took an OA first (38), and one PhD candidate's interviewer opened with a research chat (30).
  • Asked recently: graph problems in 5 of the 14 reports (29, 30, 36, 37, 38), DP or greedy in 4 (32, 35, 37, 40), heaps, intervals and ordered state in 4 (31, 32, 34, 39) and stream or sampling tasks in 2 (30, 41).
  • Difficulty: one candidate put both rounds around LeetCode medium (33); two PhD candidates said the problems felt like blends or rewrites of known problems rather than copies (34, 36).
  • Practise first: Find All People With the Secret, Directed Graph Reachability Queries, Meeting Rooms II and Largest K-Digit Subsequence.

Coverage: 14 Google SWE, SDE and PhD SWE intern candidates (reports 29–42), live technical screens and virtual onsites, report dates January 19 to February 28, 2026. Locations named: Bucharest, Hyderabad, Boston and Shanghai. Every date below is a report date.

In this series

  • Part 1: the overview, four OA reports and reports 05–12.
  • Part 2: reports 13–28.
  • Part 3 (current): reports 29–42, January–February 2026.

What came up

Graph work showed up most in this part, from plain connected components to grouping meetings that happen at the same time. Counts are out of the 14 reports here; one candidate can sit in several rows.

Topic Reports Last seen
Graphs (BFS, DFS, union-find, cycles) 5 of 14 Feb 2026
DP and greedy 4 of 14 Feb 2026
Heaps, intervals and ordered state 4 of 14 Feb 2026
Hash maps and two pointers 3 of 14 Feb 2026
Streams and random sampling 2 of 14 Feb 2026
Strings and dictionaries 2 of 14 Feb 2026
Binary search 1 of 14 Feb 2026
Research discussion (PhD) 1 of 14 Jan 2026

Questions we found

Each question below came from one candidate in this part. Report 35 also repeats a question first seen in an earlier part of the series.

Graphs

Question Reports Last seen
Course Schedule Cycle Detection 1 Feb 2026
Find All People With the Secret 1 Feb 2026
Find the Safest Path in a Grid 1 Feb 2026
Directed Graph Reachability Queries 1 Jan 2026
Number of Connected Components in an Undirected Graph 1 Jan 2026

Arrays, strings and DP

Question Reports Last seen
All Long Subsequences Are Dictionary Words 1 Feb 2026
Largest K-Digit Subsequence 1 Feb 2026
House Robber With Selected Indices 1 Feb 2026
Maximum Subarray Sum With Equal Endpoints 1 Feb 2026
Minimum Dictionary Segments (reconstruction) 1 Feb 2026
Find Minimum in Rotated Sorted Array (LeetCode 153) 1 Feb 2026
Pivot Index After Exactly One Removal 1 Feb 2026
Two Sum (LeetCode 1) 1 Jan 2026

Heaps, intervals, streams and sampling

Question Reports Last seen
Select First or Last K Stream Words 1 Feb 2026
Frequency-Weighted Next-Word Sampling 1 Feb 2026
FizzBuzz FIFO Market Matching 1 Feb 2026
Top K Frequent Closest Points 1 Feb 2026
Closest-Timestamp Key-Value Queries 1 Feb 2026
Meeting Rooms II (LeetCode 253) 1 Feb 2026
Minimum Servers for Cyclic Daily Tasks 1 Feb 2026
Reservoir Sampling With Recorded Draws (reconstruction) 1 Jan 2026

14 detailed candidate timelines, loops, and screens

Newest first. The numbers continue from Part 2.

42: Every long subsequence must be a word

Report date: February 28, 2026 · Role: SWE Intern · Level: Bachelor's · Location: Shanghai

  1. The first question asked whether every subsequence of length three or more is a real English word: All Long Subsequences Are Dictionary Words.
  2. The follow-up let each subsequence be rearranged before the dictionary check, so the answer turned into counting letters instead of matching exact order.

The interview used an open-ended "English dictionary". Our practice version gives a fixed word list so the code can run.


41: From a word list to a stream, then weighted sampling

Report date: February 28, 2026 · Role: PhD SWE Intern

  1. Select First or Last K Stream Words grew in steps: first a plain list, then a stream, then only distinct words from the stream.
  2. Frequency-Weighted Next-Word Sampling: count which word follows which in training sequences, then pick a next word with probability matching those counts. The interviewer also asked what to do when one word has a very large number of possible followers.

In both tasks the core structure stayed the same while the interviewer changed memory, order or scale.


40: Keep the largest k digits

Report date: February 2026 · Role: SWE Intern

  1. Largest K-Digit Subsequence: from a digit string, keep exactly k digits in their original order so the result is as large as possible. It is a cousin of LeetCode 402 (Remove K Digits). The usual tool is a decreasing stack that only pops while enough digits remain to reach k.

39: A playful market story that is really bookkeeping

Report date: February 18, 2026 · Role: SWE Intern · Location: Boston office · Result: No offer

  1. FizzBuzz FIFO Market Matching: Fizz and Buzz offers arrive with priorities. Opposite types match first-in, first-out; a partly filled quantity carries forward; and repeated pairs of participants must be added together in the output.

The hard part was keeping state straight, not finding a clever algorithm.


38: OA, then cycle detection over Google Meet

Report date: February 17, 2026 · Role: SWE Intern · Location: Hyderabad · Result: No offer

  1. The process was an OA followed by two technical rounds on Google Meet.
  2. The question the candidate kept was Course Schedule Cycle Detection: decide whether a course plan is possible by finding a cycle in a directed graph.
  3. Looking back, the candidate said the interviewers cared about clean code, efficiency, edge cases and optimization, not only a correct yes or no.

37: Return the chosen houses, then group meetings by time

Report date: February 14, 2026 · Role: PhD SWE Intern · Format: Two 45-minute rounds

  1. House Robber With Selected Indices: the classic House Robber, except the answer must include one best set of house indices, not just the total. The candidate finished it.
  2. Find All People With the Secret: meetings that share a timestamp have to be handled as one group. The candidate got stuck on that grouping, asked for a hint, and then talked through union-find or graph traversal.

36: An interviewer-written grid problem

Report date: February 8, 2026 · Role: PhD SWE Intern

  1. Two interviews ran back to back. The first problem was not shared.
  2. The second was confirmed as Find the Safest Path in a Grid.

The candidate stressed that the questions did not feel lifted from LeetCode, even if a familiar idea was underneath.


35: Settling an edge case from an earlier report

Report date: February 7, 2026 · Role: Software Engineer Intern · Result: No offer

  1. This candidate got Maximum Subarray Sum With Equal Endpoints and stated the rule clearly: the start index may equal the end index (i <= j), so a single element counts as a subarray. That matters when every number is negative, and it cleared up a gap in an earlier report of the same question.

34: Two blended questions

Report date: February 6, 2026 · Role: PhD SWE Intern

  1. Round one mixed Top K Frequent Elements with K Closest Points: Top K Frequent Closest Points.
  2. Round two joined a time-based key-value store with a nearest-timestamp lookup: Closest-Timestamp Key-Value Queries.

The candidate took direct approaches and discussed complexity and possible improvements with both interviewers.


33: Binary search in one round, hashing in the other

Report date: February 2, 2026 · Role: Software Engineer Intern · Location: Bucharest · Result: No offer

  1. Two back-to-back 45-minute data-structures rounds, both around LeetCode medium.
  2. One was a binary-search problem the candidate named: Find Minimum in Rotated Sorted Array.
  3. The other used hashing and two pointers; the full task was not shared.

32: Meeting rooms and a string task with no behavioral questions

Report date: February 2, 2026 · Role: SDE Intern · Level: Master's

  1. Two rounds back to back, with no behavioral questions at all.
  2. One task was described as Meeting Rooms II.
  3. The other was only remembered as string processing that recursion or DP could solve. Minimum Dictionary Segments is our practice reconstruction, not the exact interview prompt.

31: Tasks that cross midnight, then a pivot index with one removal

Report date: February 1, 2026 · Role: SWE Intern · Level: Master's · Result: No offer

  1. Round one: Minimum Servers for Cyclic Daily Tasks. Tasks start anywhere in a 24-hour cycle and can run past midnight. The candidate solved it smoothly.
  2. Round two: Pivot Index After Exactly One Removal. The candidate only reached brute force and then described an optimization out loud.

30: Research talk, reachability and sampling, then team match

Report date: January 27, 2026 · Role: PhD SWE Intern · Result: Passed to team matching

  1. The interviewer spent about 20 minutes on the candidate's research before any code.
  2. Directed Graph Reachability Queries: build adjacency lists for the relationships and answer reachability with BFS, followed by several extensions.
  3. The second task was remembered only as an applied sampling program. Reservoir Sampling With Recorded Draws is our reconstruction of it.
  4. Afterwards the candidate spoke with two teams and matched with an Ads AI group doing reinforcement-learning research.

29: A standard 45-minute screen

Report date: January 19, 2026 · Role: SWE Intern · Result: No offer

  1. A short introduction and resume chat, live coding, then time for the candidate's questions.
  2. The coding was Two Sum and Number of Connected Components in an Undirected Graph, explained and implemented inside 45 minutes.

Patterns across the reports

  1. Two coding rounds in one sitting. Reports 31–34 and 36–38 describe two technical rounds; 32, 33 and 36 say back to back, and 33 and 37 give 45 minutes each. Report 29 was a single 45-minute screen with two questions.
  2. Familiar problems with a twist. Reports 34, 36 and 37 all got a known problem with an added rule: a blend of two LeetCode tasks (34), an interviewer-written grid task (36), and House Robber that must return the chosen indices (37).
  3. Follow-ups change the rules. In 41 the input moved from a list to a stream to distinct words; in 42 the follow-up allowed rearranging letters; in 30 the reachability task had several extensions.
  4. Getting stuck on the second task. Reports 31 and 37 both finished round one but stalled on round two (brute force only in 31, a hint needed in 37).

Common questions

How many coding rounds are in the Google SWE intern interview?

Seven of the 14 reports describe two technical rounds (31–34, 36–38), 45 minutes each where a length is given (33, 37). Report 29 was one 45-minute screen.

Does the Google intern interview include behavioral questions?

Report 32 said its two rounds had none; report 29 had a short resume chat, and the PhD candidate in 30 discussed research for about 20 minutes.

Are Google intern questions taken straight from LeetCode?

Some are (Two Sum in 29, Meeting Rooms II in 32), but two PhD candidates said theirs were blends or rewrites of known problems (34, 36).

How hard are the Google SWE intern coding rounds?

Report 33 called both rounds about LeetCode medium; reports 31 and 37 each solved one round and struggled on the other.

What to practise

Keep practicing

In this series: Part 1 · Part 2 · Part 3 (current)

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.