Google Software Engineer Interview Questions

1.Tell me about a project where the requirements kept changing after you'd already started building. How did you stay productive without a stable target?

2.Walk me through the last time you had to get up to speed quickly on a codebase, system, or domain you'd never touched before. What was your approach, and how long did it take before you could contribute meaningfully?

3.Tell me about a time you were asked to estimate how long something would take, and you got it significantly wrong. How did you handle the fallout, and how do you estimate now?

4.Tell me about a time when you had to push back on a decision you disagreed with. How did you handle it?

5.Describe a project where you had to collaborate with people from different teams or disciplines. How did you make it successful?

6.What's the area of your codebase or technical stack where you'd say you have genuinely deep expertise? Give me a specific example of a problem that required that depth — something a less experienced engineer would have missed.

7.What's the most counterintuitive performance or reliability finding you've uncovered in your work — something where profiling or measurement told you something you didn't expect? What did you do with that information?

8.Tell me about a time you saw something important getting dropped — a reliability risk, a design flaw, a process gap — that wasn't your responsibility to fix. What did you do?

9.Describe a project you shipped that, in hindsight, you'd do completely differently. Not a post-launch bug — a fundamental approach you'd change. What would you do differently and why didn't you do it that way originally?

10.You're writing a Protocol Buffer schema for a new API that multiple teams will depend on. What do you think about before finalizing those field definitions?

11.You're adding a new field to a Bigtable row key that millions of rows already use. Walk me through how you'd plan and execute that migration without downtime.

12.Your service runs on Borg and you notice that job instances are being evicted far more frequently than usual. Where do you look first, and how do you reason through whether it's your problem or the infrastructure's?

13.Imagine a new Chrome feature ships and we see a sudden spike in crash reports from a specific subset of users — about 2% of the user base. The crash is reproducible but only on certain hardware configurations. How do you isolate the root cause?

14.You need to store user activity events for YouTube — things like video views, likes, and watch time — in a way that supports both real-time dashboards and long-running analytics queries. How would you architect the storage layer?

15.You're building a new internal service that several other Google teams will call via gRPC. A teammate suggests skipping versioning for now since it's just internal. How do you think about that decision?

16.A team using your gRPC service starts complaining that their tail latency — specifically P99 — is unacceptably high, even though your P50 looks fine. Walk me through how you'd diagnose this and what the likely culprits are.

17.Android's battery stats show that a popular third-party app is causing significantly higher drain on Pixel devices running the latest OS version compared to the previous version. You own the relevant framework API this app uses. How do you triage this?

18.Design an autocomplete system for Google Search. What data structures and algorithms would you use?

19.How would you design a web crawler that can index billions of web pages?

20.You're redesigning how Google Maps stores and serves road network graph data — billions of nodes and edges — for turn-by-turn routing. The current system works but doesn't scale to our latency requirements as we add more real-time data like live traffic and road closures. What's your approach?

21.You're the tech lead for a team migrating a high-QPS serving backend from hand-rolled RPC to gRPC and Protocol Buffers. Three weeks before the target cutover date, you discover that the existing binary format used by legacy clients in the field — on Android devices that haven't updated — can't be cleanly mapped to your proto schema. What do you do?

22.You've written a design doc for a new feature and a senior engineer on a critical dependency team has left a blocking comment saying your approach will cause unacceptable write amplification on their Bigtable cluster. You believe the impact is overstated. How do you resolve this?

23.Google Search indexes hundreds of billions of documents, and we're adding a new document attribute — say, 'content_language' — that we want to use as a ranking signal. The attribute doesn't exist in the index today. What's the end-to-end plan to get this signal into serving without degrading query latency or index freshness?

24.Gmail needs to support a 'Smart Reply' feature that suggests three short replies for every email a user opens. How would you design the serving layer so suggestions appear before the user finishes reading the email?

25.Walk me through how you'd design the backend for Google Maps' 'Save to List' feature — where users can bookmark places into named collections like 'Want to go' or 'Favorite coffee shops' and access them across devices.

26.Design a system that lets Chrome sync a user's open tabs, bookmarks, and browsing history across all their devices in near real-time. What does the data model and sync protocol look like?

27.Google Photos needs to send a push notification to a user when their 'Memories' are ready — a daily digest of photos from years past. How would you design the notification scheduling and delivery system for 1 billion users across Android and iOS?

28.YouTube wants to add a 'chapter' feature where creators can define named segments in their videos, and viewers can jump directly to a chapter. How would you design the backend to store and serve chapter data at YouTube's scale?

29.Design the search-as-you-type suggestions that appear in the Chrome address bar — the Omnibox — pulling from your local browsing history, bookmarks, and open tabs alongside Google Search suggestions. The whole thing has to feel instantaneous.

30.Design the system that computes and serves the 'Trending' video feed on YouTube's home page — the videos that are surging in view velocity right now. What does the pipeline look like and how do you balance freshness with accuracy?

31.Design a read receipt system for Gmail — where senders can see whether a recipient has opened their email. This is a feature enterprise Workspace customers have been asking for. Walk me through your design.

32.Design Google Cloud's managed secret storage service — something like Secret Manager — that stores API keys, passwords, and TLS certs for enterprise customers and makes them available to workloads running on GCP. What are the hardest parts of this design?

33.Design the system that powers Google Search's 'People Also Ask' box — those expandable accordion questions that appear inline in search results and dynamically load answers when you click them. Focus on how the questions are generated, stored, and served.

34.Design the system that powers live traffic overlays in Google Maps — the colored road segments showing congestion levels that update in near real-time. You have hundreds of millions of active users who are both consumers and sensors.

35.YouTube is launching a 'Clips' feature where any user can cut a 5-to-60-second clip from any public video and share it as a standalone link. The clip doesn't create a new video file — it just references a time range in the original. How do you design the end-to-end system, from clip creation to playback?

36.Tell me about a time you disagreed with a direction in your design doc review — not a minor nit, but a comment that would significantly change your approach. How did you decide whether to push back or update the doc?

37.You're midway through a project and you realize the technical approach you proposed in your design doc isn't going to meet the latency requirements you committed to. You're two weeks from a launch date. What do you do?

38.Estimate how many search queries Google processes per second on a typical weekday. Walk me through your reasoning.

39.Estimate how many active Gmail users open their inbox on a given Tuesday versus a given Saturday. What's the ratio, and how do you get there?

40.Google Maps shows an ETA to every driver. If our ETAs are systematically 15% too optimistic on highway segments during rush hour, what's the downstream impact and how would you detect it?

41.You have a dataset with millions of records and need to find the most frequent elements. What approaches would you consider?

42.YouTube's 7-day retention rate — the percentage of new users who come back within a week — dropped 3 points after a rollout of a new recommendation model. The model A/B tested well on watch time. What's going on, and what do you do?

43.Chrome has a 65% market share. Estimate the total number of tab-open events Chrome processes globally per day, and then tell me what that implies for any feature that runs synchronously on every tab open.

44.Estimate the total storage cost Google incurs annually to store Gmail messages for all users. Assume a rough cost per GB — say $0.02/GB/month. What's your number?

45.Google Search click-through rate on position-one results has dropped 8% over the past quarter, but overall query volume is up 12%. Revenue is flat. Walk me through what's happening and what you'd investigate.

46.Estimate the total annual cost to Google of storing all YouTube thumbnails — just the thumbnails, not the video files. Show your work.

47.Android has a 3 billion device install base. You're designing a feature flag system that lets Google gradually roll out a change to the Android framework itself — not an app, the OS framework — to 0.1%, then 1%, then 10%, then 100% of devices. What makes this fundamentally different from rolling out a server-side change, and what does that force you to design differently?