Google Software Engineer Interview Process & Questions (2026)
Updated every weekLast refreshed 9 Nov 2026Next refresh 16 Nov 2026
In the last 30 days, 2 candidates posted about their Google interview. This page has 40 reports so far; 14% of those with a known location are from the US. The rounds they mention most: technical interview (33), behavioral (23), onsite (15). The question that comes up most: “Find the Safest Path in a Grid”, reported 2 times in 90 days.
Google runs a different software engineer loop at each level: L3 (SWE II) has 4 rounds; L4 (SWE III) has 4 rounds; L5 (Senior SWE) has 5 rounds. Each level below lists every round with its length and format, whether your code is run, and the AI policy.
Rehearse the Google interview before the real one
The AI interviewer is set up for Google. It asks a question, pushes on your answers with follow-ups, and grades you at the end, so you know what to fix.
Start a mock interview for GoogleGoogle interview process by level
Verified loops: every round with its length and format, and whether your code is run.
- Google L3 (SWE II) · 4 rounds
- Google L4 (SWE III) · 4 rounds
- Google L5 (Senior SWE) · 5 rounds
Can you use AI?
Google does not allow AI to help answer questions during its interviews, and using it results in disqualification. Some US software engineering teams are piloting an AI-assisted code comprehension round with Gemini. Compare with other companies
Guides for the Google interview
Recently reported Google interview questions
Pulled from the reports below and updated every week. When a question matches a problem we have, you can practise it in the code editor.
Given an N by N grid of land and water cells, find a path for a rat from source to target moving in four directions that maximizes the minimum Manhattan distance to a stationary cat across visited cells; return -1 if no path exists.
Phone screen · Coding · reported 2 times in 90 days · last seen 3 Oct 2026
Practise this problemMatches Find the Safest Path in a Grid. Open it in the code editor and run it against test cases.
Process a live stream of timestamped messages and show a message only if the same text does not appear again within a 10-second window; when duplicates occur within that window, suppress both. Output must be delayed until safe, without a fixed 10-second delay, and with bounded memory.
Technical interview · Coding · reported 2 times in 90 days · last seen 4 Sept 2026
Practise this problemMatches Logger Rate Limiter. Open it in the code editor and run it against test cases.
Given N cities and M roads, each road costs 1 fuel. A car starts at the source with P fuel and must refuel to exactly P at stations. Never go below zero fuel; minimize stations visited, then roads used. Roads can be reused. Return {min_stations, min_roads} or report that it is impossible.
Technical interview · Coding · last seen 27 Oct 2026
How do you respond when a colleague claims credit for your work?
Behavioral · Behavioral · last seen 23 Oct 2026
What do you do when you are working on a project and have no idea how to proceed?
Behavioral · Behavioral · last seen 23 Oct 2026
Describe a career disappointment you have experienced.
Behavioral · Behavioral · last seen 23 Oct 2026
Using a single hardware timer that can have one pending timeout, design a timer structure and implement set_timer(t, relative_timeout_ms, callback) to support arbitrarily many concurrent timers, plus the handle_hw_timer interrupt handler. Callbacks run in interrupt context; a get_current_time function returns milliseconds since boot.
Technical interview · Coding · last seen 23 Oct 2026
Define various AI terms, with focus on prompt injection and system instructions and how they affect AI behavior.
Technical interview · last seen 22 Oct 2026
Given an array, find the maximum subarray score where score equals first element plus last element minus the sum of the middle elements. For [3,1,2,4,5], the best is [4,5] giving 9.
Onsite · Coding · last seen 22 Oct 2026
Given a connected acyclic graph where each node has at most three neighbours, find a node that can serve as the root of a binary tree.
Technical interview · Coding · last seen 20 Oct 2026
Extend the problem so each node is black or white and colors must alternate at every depth of the tree; find a valid root node.
Technical interview · Coding · last seen 20 Oct 2026
Extend the problem so the graph may be disconnected, and find a valid root.
Technical interview · Coding · last seen 20 Oct 2026
Nodes are red, white or black; depths must follow the repeating color sequence R, W, B, and the root may be any color. Find a valid root efficiently, beyond brute force BFS/DFS from every node.
Technical interview · Coding · last seen 20 Oct 2026
Describe how you built services listed on your resume and the tradeoffs involved in those designs.
Behavioral · Behavioral · last seen 14 Oct 2026
Given a graph of nodes and transmission-line edges with efficiency values between 0 and 1, find the path from a source to a destination with maximum efficiency.
Technical interview · Coding · last seen 14 Oct 2026
Practise this problemMatches Path with Maximum Probability. Open it in the code editor and run it against test cases.
Given an N by M satellite grid where X marks land and dots mark water, with a coordinate provided, use BFS to explore the island, similar to counting islands, and implement it.
Technical interview · Coding · last seen 9 Oct 2026
Practise this problemMatches Number of Islands. Open it in the code editor and run it against test cases.
Follow-up: identify or count lakes enclosed by islands in the grid.
Technical interview · Coding · last seen 9 Oct 2026
How would you handle a project that is unclear or has ambiguity?
Behavioral · Behavioral · last seen 9 Oct 2026
Explain when you use AI-generated code, when you avoid it, and how you verify it.
Phone screen · Behavioral · last seen 5 Oct 2026
Given N points in a plane, count the number of squares that can be formed, including rotated squares.
Phone screen · Coding · last seen 5 Oct 2026
Follow-up: handle duplicate points in the square counting problem, considering whether the helper function that returns the other two diagonal points suffices.
Phone screen · Coding · last seen 5 Oct 2026
Follow-up: change the problem to counting rhombi among the given points.
Phone screen · Coding · last seen 5 Oct 2026
What is the most challenging project you have worked on, and how would you approach a similar problem in the future?
Behavioral · Behavioral · last seen 3 Oct 2026
Describe a positive leadership or management style from a previous manager and how it influenced your work style.
Behavioral · Behavioral · last seen 3 Oct 2026
Tell me about a time you advocated for yourself or a teammate.
Behavioral · Behavioral · last seen 3 Oct 2026
Describe a project you proposed and led, and how you carried it out.
Behavioral · Behavioral · last seen 3 Oct 2026
What new learnings will you apply to perform better in the future?
Behavioral · Behavioral · last seen 3 Oct 2026
Given an array of numbers and a target, find two elements that sum to the target; discuss brute force, sorting with two pointers, and hashmap approaches, plus edge cases and overflow.
Technical interview · Coding · last seen 2 Oct 2026
Practise this problemMatches Two Sum. Open it in the code editor and run it against test cases.
One function returns the 4 best ads for a user and another returns the worst ad from a given set. Find the best ad, then with only one call to each function, maximize the probability of returning the best ad.
Technical interview · last seen 2 Oct 2026
Describe an ambiguous situation you handled.
Behavioral · Behavioral · last seen 2 Oct 2026
Talk about a time you received negative feedback.
Behavioral · Behavioral · last seen 2 Oct 2026
How do you manage work with remote employees?
Behavioral · Behavioral · last seen 2 Oct 2026
Variation of the maximum good subarray sum problem: return any pair of indices i and j for a good subarray instead of the maximum sum.
Technical interview · Coding · last seen 27 Sept 2026
Practise this problemMatches Maximum Good Subarray Sum. Open it in the code editor and run it against test cases.
Given a string of only 'a' and 'b' with even counts of each, make at most two cuts and split pieces between two people so each gets equal counts of 'a' and of 'b'. Return the cut indices.
Onsite · Coding · last seen 27 Sept 2026
After removing any one element from an integer array, determine whether the remaining array can be split into two parts with equal sums.
Onsite · Coding · last seen 27 Sept 2026
Simple behavioral and hypothetical situation questions.
Behavioral · Behavioral · last seen 27 Sept 2026
In a W by H grid, starting at (0,0) and ending at (w-1,0), with moves (x+1,y), (x+1,y+1), (x+1,y-1), count the number of distinct ways to reach the destination.
Phone screen · Coding · last seen 23 Sept 2026
Using the same grid and moves, count the number of ways to reach the destination that pass through a given cell (a,b).
Phone screen · Coding · last seen 23 Sept 2026
An unnamed algorithm problem with an optimal linear-time solution, followed by a harder follow-up requiring pseudocode and time and space complexity analysis.
Technical interview · Coding · last seen 22 Sept 2026
Medium-difficulty coding problem on finding a maximum/minimum value in lists, solvable using binary search.
Online assessment · Coding · last seen 20 Sept 2026
Recent Google interview experiences
Newest first. Each summary is ours; the full story is in the original post.
Earlier in 2026
Forward Deployed Engineer
Onsite, Technical interview · India · 27 Oct 2026
The candidate describes an onsite loop for a Forward Deployed Engineer position at Google in India, mentioning a compensation figure of about 95 lakhs. They say the data structures and algorithms round went poorly. The post shares the graph problem they were given, involving a network of cities connected by roads, a vehicle with limited fuel, and refueling stops, where the goal is to reach a destination while optimizing for stops first and then road count. No final outcome is stated.
- Given N cities and M roads, each road costs 1 fuel. A car starts at the source with P fuel and must refuel to exactly P at stations. Never go below zero fuel; minimize stations visited, then roads used. Roads can be reused. Return {min_stations, min_roads} or report that it is impossible.
Read the original on LeetCode Discuss →The interviewer asks the question from this report.
Run this Google interview with the AI interviewerSoftware Engineer · New grad
Behavioral, Technical interview · 23 Oct 2026
The candidate describes a Google L3 interview consisting of two parts. The first was a behavioral conversation about teamwork and leadership, covering topics such as handling someone claiming your achievements, coping with uncertainty on a project, and discussing a professional letdown. The second was a coding session with a single embedded systems problem: building a multi-timer facility on top of one hardware timer, including defining a structure, a function to schedule timers with callbacks, and the interrupt handler. No outcome was reported.
- How do you respond when a colleague claims credit for your work?
- What do you do when you are working on a project and have no idea how to proceed?
- Describe a career disappointment you have experienced.
- Using a single hardware timer that can have one pending timeout, design a timer structure and implement set_timer(t, relative_timeout_ms, callback) to support arbitrarily many concurrent timers, plus the handle_hw_timer interrupt handler. Callbacks run in interrupt context; a get_current_time function returns milliseconds since boot.
Read the original on LeetCode Discuss →The interviewer asks the 4 questions from this report.
Run this Google interview with the AI interviewerCloud AI Engineer
Technical interview, Onsite · 22 Oct 2026 · Rejected
The candidate interviewed for a Cloud AI engineering position at Google. One session tested conceptual knowledge of AI terminology, with emphasis on prompt injection and system instructions. The onsite coding round had a single array problem involving a custom subarray score. The candidate solved it with a running-sum approach and mentioned a prefix-sum optimization, but was rejected because of bugs in the code.
- Define various AI terms, with focus on prompt injection and system instructions and how they affect AI behavior.
- Given an array, find the maximum subarray score where score equals first element plus last element minus the sum of the middle elements. For [3,1,2,4,5], the best is [4,5] giving 9.
Read the original on LeetCode Discuss →The interviewer asks the 2 questions from this report.
Run this Google interview with the AI interviewerSoftware Engineer · Mid-level
Technical interview · 20 Oct 2026 · Rejected
The candidate, a software engineer with three years of experience, interviewed for an L4 position at Google and described the first DSA round. A single graph problem was extended through three follow-ups, each allotted about ten minutes. The first three parts were solved and coded, but the last stalled on optimization with little interviewer guidance. The candidate was rejected, later rounds were cancelled, and the experience was described as frustrating.
- Given a connected acyclic graph where each node has at most three neighbours, find a node that can serve as the root of a binary tree.
- Extend the problem so each node is black or white and colors must alternate at every depth of the tree; find a valid root node.
- Extend the problem so the graph may be disconnected, and find a valid root.
- Nodes are red, white or black; depths must follow the repeating color sequence R, W, B, and the root may be any color. Find a valid root efficiently, beyond brute force BFS/DFS from every node.
Read the original on LeetCode Discuss →The interviewer asks the 4 questions from this report.
Run this Google interview with the AI interviewerSoftware Engineer · Senior
Phone screen, Behavioral, Technical interview · USA · 14 Oct 2026 · Rejected
The candidate described a two-part Google phone screen for an SDE 3 role in the USA. The first part was a resume-focused discussion about building services and the tradeoffs involved. The second was a graph problem about finding the most efficient route. They proposed Dijkstra's algorithm, struggled when pressed on BFS, ran out of time while coding, and skipped clarifying constraints. They later received a rejection email.
- Describe how you built services listed on your resume and the tradeoffs involved in those designs.
- Given a graph of nodes and transmission-line edges with efficiency values between 0 and 1, find the path from a source to a destination with maximum efficiency. (Practise this problem)
Read the original on LeetCode Discuss →The interviewer asks the question from this report.
Run this Google interview with the AI interviewerSoftware Engineer · Mid-level
Online assessment, Technical interview, Behavioral · 9 Oct 2026 · Rejected
The candidate interviewed for a Google L4 software engineering role, passing an initial assessment before a virtual technical round and a behavioral round. The technical session involved a grid flood-fill problem with a confusing follow-up about enclosed lakes, which the candidate could not clarify. The behavioral round included a question on handling ambiguity that was answered poorly. The result was a rejection, with feedback to strengthen technical fundamentals and practice more problems.
- Given an N by M satellite grid where X marks land and dots mark water, with a coordinate provided, use BFS to explore the island, similar to counting islands, and implement it. (Practise this problem)
- Follow-up: identify or count lakes enclosed by islands in the grid.
- How would you handle a project that is unclear or has ambiguity?
Read the original on LeetCode Discuss →The interviewer asks the 2 questions from this report.
Run this Google interview with the AI interviewerSoftware Engineer · Mid-level
Onsite, Technical interview, Behavioral · 8 Oct 2026 · Waiting to hear back
The candidate finished a full Google onsite loop for an L4 position and is awaiting a decision before team matching. They self-assessed each round: two coding rounds went strongly, including follow-ups solved without help, one coding round needed a couple of hints, and the behavioral round had some mistakes. They ask the community to estimate their odds under the current hiring bar.
Read the original on LeetCode Discuss →Software Engineer · New grad
Phone screen, Technical interview, Behavioral · 5 Oct 2026 · Waiting to hear back
The candidate describes a roughly 45-minute Google L3 phone screen that opened with a few questions on how they use AI when coding and verify its output. The main task was counting squares, including rotated ones, from a set of points, followed by follow-ups on duplicate points and counting rhombi. They needed a hint to reach the simpler approach and did not finish the rhombus part, and they ask about their onsite chances.
- Explain when you use AI-generated code, when you avoid it, and how you verify it.
- Given N points in a plane, count the number of squares that can be formed, including rotated squares.
- Follow-up: handle duplicate points in the square counting problem, considering whether the helper function that returns the other two diagonal points suffices.
- Follow-up: change the problem to counting rhombi among the given points.
Read the original on LeetCode Discuss →The interviewer asks the 4 questions from this report.
Run this Google interview with the AI interviewerCandidate report
Phone screen, Behavioral · 3 Oct 2026
The candidate describes a virtual Google interview process from August 2026. The screening round featured a grid pathfinding problem in which a rat must travel from a source to a target while staying as far as possible from a stationary cat, followed by a question about handling several cats. The second round was a behavioral Googliness interview covering past projects, leadership influences, self-advocacy and personal growth.
- Given an N by N grid of land and water cells, find a path for a rat from source to target moving in four directions that maximizes the minimum Manhattan distance to a stationary cat across visited cells; return -1 if no path exists. (Practise this problem)
- How would you handle the safest path problem if there are multiple cats? (Practise this problem)
- What is the most challenging project you have worked on, and how would you approach a similar problem in the future?
- Describe a positive leadership or management style from a previous manager and how it influenced your work style.
- Tell me about a time you advocated for yourself or a teammate.
- Describe a project you proposed and led, and how you carried it out.
- What new learnings will you apply to perform better in the future?
Read the original on LeetCode Discuss →The interviewer asks the 5 questions from this report.
Run this Google interview with the AI interviewerSoftware Engineer · Mid-level
Phone screen, Online assessment, Behavioral, Technical interview · India · 2 Oct 2026 · Rejected
The candidate, a software engineer with about three years of experience, contacted a Google recruiter and went through a recruiter call, then online assessments covering behavioral scenarios and two DSA problems. A behavioral interview went very well, but the DSA interview began with a simple array problem and moved on to a probability puzzle about ad selection functions. The candidate could not find the optimal strategy and was rejected with a lean no hire.
- Given an array of numbers and a target, find two elements that sum to the target; discuss brute force, sorting with two pointers, and hashmap approaches, plus edge cases and overflow. (Practise this problem)
- One function returns the 4 best ads for a user and another returns the worst ad from a given set. Find the best ad, then with only one call to each function, maximize the probability of returning the best ad.
- Describe an ambiguous situation you handled.
- Talk about a time you received negative feedback.
- How do you manage work with remote employees?
Read the original on LeetCode Discuss →The interviewer asks the 4 questions from this report.
Run this Google interview with the AI interviewerSoftware Engineer · New grad
Technical interview, Behavioral, Onsite · India · 27 Sept 2026 · Got an offer
The candidate describes a Google India L3 software engineer process with four interviews: one virtual coding round, one virtual behavioral round, and two back-to-back onsite coding rounds. All coding problems were solved, with feedback to improve time management. Team matching followed within days, then hiring committee review, compensation discussion and an offer, which was accepted. The candidate also mentions many earlier unsuccessful applications to the company.
- Variation of the maximum good subarray sum problem: return any pair of indices i and j for a good subarray instead of the maximum sum. (Practise this problem)
- Given a string of only 'a' and 'b' with even counts of each, make at most two cuts and split pieces between two people so each gets equal counts of 'a' and of 'b'. Return the cut indices.
- After removing any one element from an integer array, determine whether the remaining array can be split into two parts with equal sums.
- Simple behavioral and hypothetical situation questions.
Read the original on LeetCode Discuss →The interviewer asks the 3 questions from this report.
Run this Google interview with the AI interviewerSoftware Engineer · Mid-level
Phone screen · 23 Sept 2026
The candidate, with about three years of experience, describes a Google phone screen built around a grid path-counting problem. Starting at the top-left corner and aiming for the far end of the bottom row, the movement options are forward, forward-up, or forward-down. The task had two parts: count all routes to the destination, then count only routes that pass through a specified intermediate cell. No outcome is shared.
- In a W by H grid, starting at (0,0) and ending at (w-1,0), with moves (x+1,y), (x+1,y+1), (x+1,y-1), count the number of distinct ways to reach the destination.
- Using the same grid and moves, count the number of ways to reach the destination that pass through a given cell (a,b).
Read the original on LeetCode Discuss →The interviewer asks the 2 questions from this report.
Run this Google interview with the AI interviewerCandidate report
Technical interview · 22 Sept 2026
The candidate describes a Google coding interview in which they explained an approach, coded it, and the interviewer agreed it was optimal. A follow-up was also solved with pseudocode and complexity analysis. Later, the candidate verified online that their linear-time solution matched the best known one, yet the interviewer's written feedback reportedly said something different. They are unsure whether to ask the recruiter for clarification.
- An unnamed algorithm problem with an optimal linear-time solution, followed by a harder follow-up requiring pseudocode and time and space complexity analysis.
Read the original on LeetCode Discuss →The interviewer asks the question from this report.
Run this Google interview with the AI interviewerCandidate report
Online assessment · 20 Sept 2026 · Rejected
A third-year computer science student took the qualifying round of a Google coding contest on HackerEarth. The 45-minute test had 17 multiple-choice questions on fundamentals, AI and systems, followed by one medium-level coding task involving binary search on lists. The candidate spent about twenty minutes on the MCQs, then ran out of time while writing a brute-force solution, and was logged out. A rejection email arrived five days later. The advice is to attempt the coding problem first.
- Medium-difficulty coding problem on finding a maximum/minimum value in lists, solvable using binary search.
- 17 multiple-choice questions covering CS fundamentals, AI and system understanding, requiring deep subject knowledge.
Read the original on LeetCode Discuss →The interviewer asks the 2 questions from this report.
Run this Google interview with the AI interviewerSoftware Engineer II (L3) · Mid-level
Online assessment, Phone screen, Behavioral, Onsite, Technical interview · India · 18 Sept 2026 · Waiting to hear back
The candidate describes a Google L3 process in India: an online situational assessment, a coding phone screen, a behavioral phone screen, and two onsite coding rounds. Questions covered dynamic programming, graph traversal with thresholds, and tree edge-cutting with follow-ups. The write-up includes a detailed timeline with recruiter changes and rescheduled interviews. Team matching has taken place, but the final result is still pending.
- Find the length of the longest subsequence in an integer array where consecutive elements increase by exactly 1; optimize from quadratic DP to linear with a hash map.
- Follow-up: allow consecutive differences anywhere from 1 to d in the longest increasing subsequence. (Practise this problem)
- How do you handle conflict, missed deadlines, unclear goals, and a colleague repeatedly taking credit for your work?
- How do you use AI in daily work, what would you want at work that does not exist today, and what do you want to achieve soon in your career?
- In a directed graph with edge security levels, find the minimum S to travel from u to v where an edge is usable only if its level is below S; discuss negative levels, cycles, queue versus priority queue.
- Given a binary tree with positive edge weights, find the minimum total cost of cuts that remove all leaf nodes; follow-ups for N-ary trees and negative edge weights.
Read the original on LeetCode Discuss →The interviewer asks the 5 questions from this report.
Run this Google interview with the AI interviewerCandidate report
Technical interview · 16 Sept 2026
The candidate described a Google interview round featuring a geometry problem about axis-aligned squares. They were allowed to explain an approach instead of coding. They proposed a binary search over the vertical range with a linear area check, giving O(n log y) complexity. The interviewer wanted something faster, and time ran out. Afterward the candidate realized a linear solution using an area-weighted average of the squares' vertical centers.
- Given axis-aligned squares, each with a known top-left corner and edge length, find a horizontal line that splits the total combined area of all squares into two equal halves. Example: corners [(1,2),(2,4)] with sizes [5,10].
Read the original on LeetCode Discuss →The interviewer asks the question from this report.
Run this Google interview with the AI interviewerCandidate report
Technical interview · 16 Sept 2026
The candidate describes a Google coding round built around a shortest-path graph problem on cities and roads, where the goal was to find the nearest favourite city from a source. The interviewer then pushed a lengthy discussion on algorithm complexity, followed by several variations including early termination and a mandatory intermediate city. The candidate needed a little guidance but reached a working approach.
- Given cities and weighted roads, a source city and a list of favourite cities, find the favourite city reachable from the source in minimum travel time.
- Discuss the time complexity of the shortest-path solution, including the inner loop.
- Add early stopping conditions to the shortest-path search.
- Modify the solution so the route to the favourite city must pass through a specified vertex V.
Read the original on LeetCode Discuss →The interviewer asks the 4 questions from this report.
Run this Google interview with the AI interviewerSoftware Engineer · Mid-level
Phone screen, Behavioral, Onsite, Technical interview · 16 Sept 2026 · Waiting to hear back
The candidate describes a Google L4 loop in August 2026 consisting of a phone screen, a Googlyness round, and two onsite coding rounds. Problems covered grid path counting with checkpoints, shortest-path graph reasoning involving two moving people, and finding when a social network becomes fully connected, including unfriend events. Feedback was reportedly positive for each round, with the last rated higher, and the candidate is now in team matching.
- Count paths in an n x m grid from bottom-left to bottom-right using moves diagonally up, right, or diagonally down.
- Extend the grid path count to paths that visit all checkpoints, then to checkpoints visited in order, while optimizing space.
- How would you handle a conflict where someone takes credit for your work?
- Describe how you juggle multiple tasks and priorities.
- Find the shortest path between two nodes in an undirected unweighted graph, then find all nodes on any shortest path.
- Alice follows a chosen shortest path while Bob moves simultaneously from another node; determine whether Alice can reach the destination without being caught.
- Given timestamped connection logs and a user count, find the timestamp when all users become connected. (Practise this problem)
- Variation: find the time after which maximum connections between any two nodes is at most k and everyone is connected; then handle unfriend events in the logs.
Read the original on LeetCode Discuss →The interviewer asks the 7 questions from this report.
Run this Google interview with the AI interviewerSoftware Engineer III · Mid-level
Phone screen, Technical interview, Behavioral, Onsite · India · 15 Sept 2026 · Waiting to hear back
The candidate described a Google L4 software engineering process in India that began with a recruiter email and a screening call. It included a coding screen, a culture-fit screen, and two back-to-back onsite coding interviews. Problems covered graph rooting with alternating colors, counting enclosed lakes on an island with DFS, and weighted job scheduling solved with memoization and binary search. Behavioral prompts focused on feedback, conflict, and mentoring. The candidate was still waiting for a final decision at the time of writing.
- In an undirected acyclic connected graph with each node having at most 3 edges, find a node that, as root, makes the graph a binary tree. Follow-up: nodes are black or white; find a root where the tree is binary and colors alternate by depth, solved in O(N) without BFS.
- Describe a time you gave critical feedback to someone and how they reacted.
- Describe a recent challenging technical problem, how you solved it, and the result.
- How would you handle people who disagree with a majority decision on a non-work matter?
- Tell about a time your manager made reasonable demands, and a follow-up about unreasonable demands.
- Describe a disagreement with a colleague or manager, how you resolved it, and the outcome.
- Describe how you helped an underperforming teammate improve.
- Given a large grid with several islands and an API taking an island and a point on it, return how many lakes (water fully enclosed by land) that island has; ocean-connected water does not count.
- Given job start times, end times, and profits, return the maximum profit from non-overlapping jobs, in O(N log N).
Read the original on LeetCode Discuss →The interviewer asks the 9 questions from this report.
Run this Google interview with the AI interviewerSoftware Engineer · Mid-level
Technical interview, Behavioral, Onsite · 15 Sept 2026
The candidate describes a Google L4 loop with two virtual rounds followed by two onsite rounds. The first round covered grid path-finding with a follow-up on maximizing the minimum distance from a hazard, solved via search plus binary search. The second was a culture-fit conversation. Onsite included a free-form rectangle geometry problem where the code stayed incomplete, and a weighted job scheduling problem. The final result is not stated.
- Given an N by N grid with water cells, a start and a target, determine whether a path exists moving only horizontally or vertically.
- On a grid with water, find a path from start to target that maximizes the minimum Manhattan distance to a cat cell along the path. (Practise this problem)
- Why Google, and what would you do if you were a manager; questions about delivery and extra effort.
- Given a set of points, find the rectangle area where sides need not be axis-parallel (similar to the LeetCode problem).
- Given jobs with start times, end times and profits, find the maximum profit with no overlapping jobs selected.
Read the original on LeetCode Discuss →The interviewer asks the 4 questions from this report.
Run this Google interview with the AI interviewerSoftware Engineer · New grad
Technical interview · 13 Sept 2026 · Waiting to hear back
The candidate describes a Google L3 coding round built around a string puzzle with brackets and digits, where each digit forces deletions of earlier characters. The session covered clarifying edge cases, a brute-force backtracking idea, a memoized improvement, a brief mention of interval DP, and then implementation in Python with a dry run. The result is still awaited, and the author asks for a predicted verdict.
- Given a string of parentheses and digits 0-9, where each digit acts as a conditional deletion instruction for characters to its left, determine whether some valid sequence of deletions produces a strictly balanced parentheses string.
Read the original on LeetCode Discuss →The interviewer asks the question from this report.
Run this Google interview with the AI interviewerSoftware Engineer · Staff
Onsite, Technical interview · Bangalore, India · 12 Sept 2026
The candidate interviewed for a senior engineering position at Google in Bangalore and attended an in-person round. They recount one algorithmic problem involving a grid with blocked and open cells, where a closed walk of a fixed length from the starting cell must be produced, choosing the alphabetically smallest move sequence when several exist. The post also mentions a compensation figure.
- Given an N by M grid with free cells, obstacles and a start cell, find a closed walk of exactly k steps returning to the start. If several exist, output the lexicographically smallest move string using L, R, U, D.
Read the original on LeetCode Discuss →The interviewer asks the question from this report.
Run this Google interview with the AI interviewerCandidate report
Technical interview · 4 Sept 2026
The candidate describes a follow-up problem met during a technical discussion at Google, a twist on the classic logger rate limiter task. Messages arrive as a live stream, and any message repeated within ten seconds must be hidden along with its duplicate. Because later arrivals can cancel earlier ones, output must be deferred. The author invites ideas on buffering without fixed delay, bounding memory, and using a deque or linked list.
- Process a live stream of timestamped messages and show a message only if the same text does not appear again within a 10-second window; when duplicates occur within that window, suppress both. Output must be delayed until safe, without a fixed 10-second delay, and with bounded memory. (Practise this problem)
Software Engineer III · Mid-level
Phone screen, Technical interview · 3 Sept 2026 · Waiting to hear back
The candidate describes a first-round technical phone interview for a Google SWE III position in the US. The session used a shared document editor with no code execution and followed a flow of introductions, one coding problem, test walkthrough, follow-ups touching on system design, and a Q&A. The problem involved filtering repeated device events in a stream, and the candidate found the experience fairly manageable.
- Given a stream of events emitted by devices with timestamps, remove duplicate occurrences of the same event that occur within 10 seconds of each other to reduce noise. (Practise this problem)
- What are the time and space complexity of your solution?
- If deploying this code to a real production environment, what bottlenecks do you expect and how would you address them?
Read the original on LeetCode Discuss →The interviewer asks the 2 questions from this report.
Run this Google interview with the AI interviewerSoftware Engineer · Mid-level
Technical interview, Behavioral, Onsite · 27 Aug 2026 · Waiting to hear back
The candidate completed two virtual rounds, one on data structures and algorithms and one behavioral, receiving positive ratings but advice to strengthen algorithms. Two onsite rounds followed. In the first, they solved two problems unaided and only outlined a brute-force approach for a third. In the second, they produced a brute-force solution with correct complexity but could not optimize it despite hints. The candidate is now asking about their chances for the L4 level.
- Problem matching LeetCode 1102 (path with maximum minimum value in a grid); candidate wrote brute force with correct complexity but could not optimize even with hints.
Read the original on LeetCode Discuss →The interviewer asks the question from this report.
Run this Google interview with the AI interviewerSoftware Engineer · New grad
Online assessment, Technical interview, Behavioral, Onsite · Bengaluru/Hyderabad, India · 27 Aug 2026 · Got an offer
The candidate applied for an entry-level engineering position at Google in India and went through a scenario-based assessment on company principles, then a virtual technical interview paired with a behavioral conversation, followed by back-to-back virtual onsite coding sessions covering graph and dynamic programming topics. Interviewers pushed on optimization, tradeoffs, and edge cases. After clearing the process, the candidate went through team matching, received one unsuitable match call, and eventually announced a successful offer, while seeking advice from others about team matching.
- Solve a graph problem, then answer follow-ups on optimization, alternative approaches, and trade-offs.
- Discuss collaboration, ownership, conflict resolution, and lessons from past experiences.
- Solve two graph problems emphasizing algorithm design, optimization, and edge cases.
- Solve a dynamic programming problem, explaining approaches, state transitions, complexity, and optimizations.
Read the original on LeetCode Discuss →The interviewer asks the 4 questions from this report.
Run this Google interview with the AI interviewerSoftware Engineer · Mid-level
Technical interview, Behavioral, Onsite · Bangalore, India · 24 Aug 2026
The candidate interviewed for an L4 position at Google's Bangalore office, passing an initial technical round and a virtual behavioral round before being invited to two onsite rounds. One onsite round involved designing a movie recommendation feature, which the candidate solved with a union-find structure. The final round was a demanding three-dimensional dynamic programming optimization problem. The candidate found the last round very intense and did not share a result.
- Given lists of similar movies, design a top-k recommendation feature that surfaces related movies when a user clicks a movie, similar to YouTube.
- Solve a three-dimensional dynamic programming optimization problem about F1 racing tyre quality score with several additional factors.
Read the original on LeetCode Discuss →The interviewer asks the 2 questions from this report.
Run this Google interview with the AI interviewerMid-level
Behavioral, Phone screen · 22 Aug 2026 · Rejected
The candidate interviewed for an L4 position at Google, completing a Googliness behavioral round and a technical phone screen. After a month, a new recruiter reported the behavioral round went well but the technical screen was weak. The candidate disagrees, saying several approaches were offered and the problem was solved on time. The outcome was a rejection with a one-year waiting period before reapplying.
Read the original on LeetCode Discuss →Software Engineer · Mid-level
Technical interview, Behavioral · India · 20 Aug 2026 · Rejected
The candidate interviewed for an entry-level L3 software engineering position at Google in India. The process included a virtual technical round centered on designing a custom map-like data structure that supports random retrieval with removal, followed by a behavioral conversation covering teamwork, deadlines, adaptability, and a few hypothetical leadership and customer scenarios. The candidate was rejected after the technical stage, and shared the questions to help others who are preparing for similar interviews.
- Design a probabilistic map data structure supporting insertion of key-value pairs and an operation that, given a key, returns and deletes a random value stored under that key if any exist.
- Introduce yourself.
- Describe a time you faced a deadline.
- Describe a time you had to change your work style.
- Describe a time you helped someone.
- Describe a time you collaborated with another team.
- If you became CTO of your current company, what changes would you make?
- If you learned customers were unhappy with your product, what would you do?
- How do you define an ideal manager?
- How do you define an ideal software engineer?
Read the original on LeetCode Discuss →The interviewer asks the 10 questions from this report.
Run this Google interview with the AI interviewerSoftware Engineer III · Mid-level
Behavioral, Technical interview, Onsite, Other round · Bangalore, India · 20 Aug 2026 · Waiting to hear back
The candidate describes a Google SWE III (L4) loop with a virtual behavioral round and a virtual array-based coding round, followed by two onsite coding rounds in Bangalore. One onsite round went very well with a follow-up solved; the other involved a multi-source BFS approach but ran short on time, with code quality noted. A team match followed, but the candidate has heard nothing for over a week and asks about timelines.
- Behavioral round on Googliness-style topics, answered using the STAR framework.
- Array-based coding problem with a follow-up that was not reached due to time spent discussing the approach.
- Coding problem with a follow-up question, both solved and fully implemented.
- Scenario-based problem solvable with multi-source BFS, wrapped in a detailed story that made the requirements confusing.
Read the original on LeetCode Discuss →The interviewer asks the 4 questions from this report.
Run this Google interview with the AI interviewerSoftware Engineer · Mid-level
Technical interview, Behavioral, Onsite · 15 Aug 2026 · Waiting to hear back
The candidate completed a four-round Google SWE-3 loop consisting of an initial coding round, a behavioral round, and two onsite coding rounds. The recruiter relayed positive feedback on the first two, noting strong debugging and culture-fit impressions. In the third round the candidate reached an optimal solution with a small hint. In the fourth, the optimal idea was found late and the code was left unfinished with bugs. The candidate seeks an assessment of their chances.
Read the original on LeetCode Discuss →Software Engineer · Mid-level
Technical interview · Bangalore, India · 11 Aug 2026 · Rejected
The candidate describes a first-round technical interview for an L4 position at Google in Bangalore that ended in rejection. The task involved a grid of heights where the goal was to determine the longest route from any starting square, with descent rules plus a special allowance to climb back up to an earlier height. The candidate proposed trying every square as a start and running a breadth-first search, storing several previous heights in each queue entry. Afterward, the candidate felt discouraged and requested similar practice problems.
- Given a 2D grid of heights, find the longest path from any starting cell without revisiting cells. Moves must go to equal or lower height, but a step up is allowed if the new height does not exceed the height two cells back in the path. The first two cells must be non-increasing.
Read the original on LeetCode Discuss →The interviewer asks the question from this report.
Run this Google interview with the AI interviewerSWE Intern · Intern
Technical interview · India · 7 Aug 2026
The candidate interviewed for a Google software engineering internship, with a stipend quoted in lakhs per month, suggesting an India-based role. The first round centered on a tree dynamic programming problem about cutting weighted edges so that all leaves are separated from the root at minimal cost. The interviewer then asked for a generalization from binary trees to trees with arbitrary numbers of children. No outcome was shared.
- Given a binary tree with positive integer edge weights, remove edges at minimum total cost so that every leaf is disconnected from the root.
- Follow-up: how would the solution change if the tree were N-ary instead of binary?
Read the original on LeetCode Discuss →The interviewer asks the 2 questions from this report.
Run this Google interview with the AI interviewerSoftware Engineer · Mid-level
Phone screen · 4 Aug 2026
The candidate describes a Google L4 phone screen centered on one problem: grouping strings whose consecutive character gaps match cyclically. They built a canonical difference-based key stored in a hash map, then handled three follow-ups: single-character strings needing separate groups, wraparound distance across the alphabet, and extending to the full ASCII range. The interview ended after the last code change. The candidate asks about callback chances.
- Group strings into buddy groups where the cyclic distance between each pair of consecutive characters is identical, returning a list of groups.
- How should single-character strings be grouped? Different characters must go to different groups, identical ones together.
- How would strings such as abc and yza be handled with wraparound differences?
- What other corner cases exist, and how would you modify the code to support special characters across the full ASCII set?
Read the original on LeetCode Discuss →The interviewer asks the 4 questions from this report.
Run this Google interview with the AI interviewerSoftware Engineer (ML) · Mid-level
System design, Behavioral, Technical interview · 4 Aug 2026 · Waiting to hear back
The candidate, with five years of experience, interviewed for an ML-focused software engineering position at the L4 level. There were four rounds: ML system design, a culture-fit round, and two coding sessions. The design and culture rounds felt strong, while the first coding round needed a hint and had minor mistakes, and in the second the candidate found the optimal approach but ran out of time. The candidate asks about the likelihood of advancing to team matching.
Read the original on LeetCode Discuss →SWE Intern · Intern
Technical interview · 4 Aug 2026 · Waiting to hear back
The candidate describes a one-hour first-round interview for a Google software engineering internship. It featured a tree-based dynamic programming problem about cutting weighted edges to separate leaves from the root, solved with a bottom-up traversal, then extended to trees with arbitrary branching with some interviewer hints. The final minutes covered questions on using AI in daily work. The interviewer was friendly and collaborative, and the candidate found it a good discussion.
- Given a binary tree with positive edge weights, remove edges at minimum total cost so that every leaf is disconnected from the root.
- Generalize the minimum-cost leaf disconnection solution to an N-ary tree with any number of children.
- How do you use AI in your regular workflow, and do you give it full ownership of your projects?
- How do you use AI for tasks like debugging?
Read the original on LeetCode Discuss →The interviewer asks the 4 questions from this report.
Run this Google interview with the AI interviewerSoftware Engineer II (Early Careers) · Mid-level
Phone screen, Technical interview, Behavioral · 3 Aug 2026 · Waiting to hear back
The candidate describes a Google SWE II loop after roughly 1.6 years of experience as an AI engineer. It began with a recruiter call, followed by three technical rounds covering Trie with hashing, graphs with a modified union-find, and dynamic programming with binary search plus generative AI talk. A final round mixed Googliness with a data structures problem and went poorly. Feedback ranged from strong hire to lean hire, and team matching was still underway.
- Solve a medium-hard coding problem combining a Trie and a HashMap, then optimize it and discuss alternatives, complexity and edge cases.
- Solve a hard graph problem using a modified Disjoint Set Union, and justify correctness, complexity and trade-offs.
- Solve a dynamic programming problem progressing from recursion to memoization to an optimal binary search solution.
- How would you leverage generative AI for automation within the Google Ads ecosystem, and what are the challenges of deploying LLM systems at scale?
- Solve a data structures and algorithms problem, plus answer Googliness behavioral and engineering collaboration questions.
Read the original on LeetCode Discuss →The interviewer asks the 5 questions from this report.
Run this Google interview with the AI interviewerSoftware Engineer · Mid-level
Phone screen, Behavioral, Onsite · 30 Jul 2026 · Waiting to hear back
The candidate, an iOS-focused engineer, describes a Google L4 software engineer loop completed in Swift. A phone screen went well, a rushed behavioral round was poorly rated but the recruiter allowed a redo that succeeded, and two onsite coding rounds earned hire and leaning hire ratings. The candidate is now awaiting team matching and asks the community about chances, the effect of an iOS background, and expected timelines.
- Given an array of start and end times, schedule tasks and return x,y coordinates for each scheduled task.
- Behavioral questions about past engineering experiences and examples, in a Googlyness-style round.
- Design a multiuser heart rate monitor, including classes, relationships, and data structures chosen to minimize message delay.
- Solve a divide and conquer problem involving a maximum height condition and simulating horizontal strokes efficiently.
Read the original on LeetCode Discuss →The interviewer asks the 4 questions from this report.
Run this Google interview with the AI interviewerWeb Solutions Engineer
Phone screen · 27 Jul 2026 · Waiting to hear back
The candidate describes a recruiter phone screen for a Web Solutions Engineer position at Google that ran about ten minutes past its scheduled half hour. The recruiter had no technical background and asked general frontend and backend questions for roughly twenty minutes, then explained the role and upcoming interview stages. Afterward the application status moved to the scheduling queue. The author advises sounding confident when speaking with the recruiter.
- What is dependency injection?
- How would you optimize the performance of a web application?
- How would you debug a website that is loading slowly?
Read the original on LeetCode Discuss →The interviewer asks the 3 questions from this report.
Run this Google interview with the AI interviewerSoftware Engineer (SDE III) · Senior
Technical interview, Behavioral · 27 Jul 2026 · Got an offer
The candidate describes a four-round Google loop for an L4 software engineering position and reports being selected. Three rounds were algorithmic: designing an ad-serving interface with score ranking and cooldown rules, building a URL router that extracts path parameters, and a tree problem about finding valid roots with alternating level colors. One round was a behavioral conversation, answered using a structured storytelling method, with follow-up probing on reasoning and impact.
- Design an ad store with InsertAd(content, score) and GetAd() returning the highest-scoring ad, decrementing its score after each return, and never returning the same ad on consecutive calls.
- Follow-up: each ad has a delay, so after being returned it cannot be returned again for the next delay GetAd calls.
- Behavioral and situational questions about collaboration, ownership, conflict resolution and decision-making.
- Implement registerHandler(route) and getRouteHandler(url) for web-framework-style route matching with path parameters like /users/{id}/pictures/{pictureId}, returning null if nothing matches.
- Follow-up: also return the extracted parameter names and values along with the matched route.
- Given an undirected acyclic graph, find all nodes that, when used as root, form a binary tree whose levels alternate in a single color each (BWBW or WBWB).
- Follow-up: find the minimum number of color changes needed to satisfy the conditions.
Read the original on LeetCode Discuss →The interviewer asks the 7 questions from this report.
Run this Google interview with the AI interviewer
Wrote one of these posts and want it taken down? Tell us and we'll remove it.
Questions, answered
- How many rounds are in the Google interview?
- It depends on the level: L3 (SWE II) has 4 rounds; L4 (SWE III) has 4 rounds; L5 (Senior SWE) has 5 rounds. Each level page lists every round with its length and format.
- What questions does Google ask in interviews?
- The ones candidates reported most recently: Find the Safest Path in a Grid (2 times in 90 days); Logger Rate Limiter (2 times in 90 days); Given N cities and M roads, each road costs 1 fuel. A car starts at the source with P fuel and must refuel to exactly P at stations. Never go below zero fuel; minimize stations visited, then roads used. Roads can be reused. Return {min_stations, min_roads} or report that it is impossible.. The full list is on this page and grows with each refresh.
- Can you use AI in the Google interview?
- Google does not allow AI to help answer questions during its interviews, and using it results in disqualification. Some US software engineering teams are piloting an AI-assisted code comprehension round with Gemini.
- How often is this page updated?
- Every 7 days. Each refresh looks for new public Google interview posts, checks the link works and adds what it finds. Old reports stay. The last refresh finished on 9 Nov 2026; the next is due on 16 Nov 2026.
- How do I practise these Google questions?
- A question with a practice button matches a problem in our catalogue. The button opens the AI interviewer on that problem with Google set as the company: it asks the question, follows up on your approach and grades you at the end.
- Is this the official Google interview process?
- No. Thita is not affiliated with Google. The loops by level are checked against current public descriptions of the process; the reports are candidates' own accounts, so treat them as what people saw recently.
Get new Google interview questions every week
When a refresh finds new Google reports, you get the questions and a link to each post. Nothing else.