Google Senior Software Engineer Interview Questions

1.Tell me about a project where the requirements kept shifting mid-execution. How did you decide when to keep pushing forward versus stop and re-plan?

2.Describe a time you were handed a completely unfamiliar codebase or system and had to diagnose a live production problem in it. Walk me through exactly how you approached it.

3.Tell me about a technical decision you made that you later regretted. What did you learn?

4.Describe a time when you led a technical initiative that required buy-in from multiple teams.

5.Tell me about a time you changed the technical direction of a project you didn't own. How did you get the team that owned it to actually change course?

6.You're adding a new field to a Protocol Buffer message that's shared across dozens of services in the monorepo. What's your process for doing this safely?

7.You're building a new backend service on GCP that needs to store user preference data — small records, high read frequency, low write frequency, with sub-10ms P99 read latency at 100K QPS. How do you choose between Bigtable, Spanner, and Firestore?

8.You're working on a Search ranking pipeline that runs as a MapReduce job and it's taking 40% longer than it did three months ago. Nothing obvious changed in the job code. How do you diagnose this?

9.We're deprecating an internal gRPC API that's used by 200+ services across the monorepo. How do you execute that migration safely?

10.YouTube's video upload pipeline needs to transcode every uploaded video into a dozen different formats and resolutions. This is currently a synchronous pipeline, and it's becoming a bottleneck at scale. How would you redesign it?

11.Your team is writing a new service that will be called by 50 other internal services. You're deciding between exposing it via gRPC with proto contracts versus a REST/JSON API. What drives your decision, and what would you tell the 50 teams?

12.YouTube needs to serve personalized video thumbnails — the same video shows a different thumbnail to different users based on predicted click-through rate. Walk me through how you'd design the serving layer for this, focusing on latency and consistency trade-offs.

13.Chrome's Safe Browsing feature checks URLs against a blocklist to protect users from phishing. The current design sends a hash prefix of each URL to Google's servers, but privacy advocates are pushing for a fully on-device check. How would you evaluate that trade-off and decide?

14.How would you design a global-scale distributed database?

15.Google Search serves billions of queries a day from a global serving infrastructure. You're tasked with designing a feature flag system that lets you roll out ranking algorithm changes to 0.1% of traffic in a specific region, with the ability to kill it in under 30 seconds if something goes wrong. Walk me through the design.

16.BigQuery (Dremel) allows users to run ad-hoc SQL queries over petabyte-scale datasets with no pre-defined indexes. A new category of enterprise customer is running queries with full-table scans on tables that have 90% of their data from the last 7 days and 10% from years of history — but their queries almost always filter on a recent time window. How do you improve their query performance without changing the query itself?

17.Google Search indexing relies on a pipeline where multiple teams own different stages — crawling, parsing, scoring, and serving. You notice that index freshness for a category of news pages has degraded by 6 hours over three months, but no single team's metrics show a problem. How do you find the root cause and drive a fix?

18.Spanner is being considered as the replacement for a legacy MySQL-based data store used by a high-QPS Google Cloud service. The migration would affect 12 teams and take 18 months. Make the case for how you'd structure this migration — and specifically, what conditions would make you recommend against doing it at all.

19.Gmail needs to support a 'Smart Reply' feature that suggests short responses to emails. The suggestions must appear in under 200ms. How do you design the serving architecture for this?

20.Google Maps needs to show real-time traffic conditions on the map — speeds, incidents, road closures. How would you design the data pipeline from raw GPS signal to rendered tile?

21.Design a real-time collaborative document editing system like Google Docs.

22.Tell me about a time you wrote a design doc that got significant pushback. How did you handle the feedback, and what did the final outcome look like?

23.Tell me about a time you sponsored or mentored someone junior on your team — not just helped them, but actively advocated for them. What did you do, and what happened to them?

24.Tell me about a time when you had to make a trade-off between shipping quickly and code quality.

25.You're 8 months into a high-visibility project when you realize the core technical assumption in the original design doc was wrong. The launch is 4 months away and two other teams have already built dependencies on your API. How do you handle it?

26.Describe a time you pushed to kill a project or feature that had organizational momentum behind it. How did you make the case, and what happened?

27.A critical internal platform your team depends on has been quietly degrading for two years — response times up, error rates creeping, docs out of date — but the team that owns it is overloaded and keeps deprioritizing it. You don't own the platform and have no authority over that team. How do you actually move this forward?

28.Android has 3 billion active devices. If Google pushes a system update that increases battery drain by 2% on average, walk me through the real-world impact of that — and how you'd decide whether to ship it.

29.Chrome holds about 65% browser market share. A proposed feature would improve page load speed by 8% for Chrome users but requires a non-standard API that other browsers can't implement for at least two years. How do you think through whether Google should ship it?

30.Google Maps has 1 billion monthly users. Suppose we discover that ETA accuracy in a specific Latin American country has been degrading for six months — estimated arrival times are off by 15 minutes on average — but no single team's health dashboard is alerting. How do you find the root cause and fix it without owning the relevant systems?

31.Suppose you discover that a shared internal Pub/Sub-based event bus used by 300+ services across Google has a subtle ordering guarantee bug — under specific high-load conditions, events arrive out of order in ways that corrupt state in downstream consumers. The bug has probably been present for over a year. How do you handle this?