Google · Onsite
RejectedGoogle ML Engineering Manager Onsite: 1 Experience
A 2026 Google ML-direction manager loop: a behavioral round, then a backend design for read-only lookups over a billion IDs; rejected the same day.
At a glance
- Process: a behavioral round one day, a system design round the next, a rejection three hours later (report 01)
- Asked recently: impact beyond your own team, an unreasonable request from a manager, a coworker's poor customer service, a coworker attached to the old way, a read-only lookup over about a billion IDs (report 01)
- Difficulty: the candidate rated it average and felt the design round went poorly
- Practise first: Design a Read-Only Lookup Service for a Billion Records
Coverage: One experienced candidate with a Master's degree, changing jobs, interviewing for a manager role in a machine learning direction. The report covers a behavioral round and a system design round. It gives October to December 2026 as the interview period and went up on November 18, 2026.
What came up
The candidate expected the loop to lean on machine learning and agents, since the role was in an ML direction. Instead the design round was a classic backend storage problem. The interviewer took the front end, and any third-party system the candidate used had to be explained from the inside.
| Topic | Reports | Last seen |
|---|---|---|
| Behavioral: influence, pushback, peers | 1 of 1 | Nov 2026 (report date) |
| System design: read-only lookup over about a billion IDs | 1 of 1 | Nov 2026 (report date) |
Questions we found
Behavioral
| Question | Reports | Last seen |
|---|---|---|
| An achievement whose impact went beyond your own team | 1 | Nov 2026 (report date) |
| Your manager made an unreasonable request | 1 | Nov 2026 (report date) |
| A coworker was giving customers poor service | 1 | Nov 2026 (report date) |
| A coworker stuck with a familiar approach | 1 | Nov 2026 (report date) |
System Design
| Question | Reports | Last seen |
|---|---|---|
| Backend for read-only lookups over about a billion IDs, tens of petabytes (practice) | 1 | Nov 2026 (report date) |
1 detailed candidate timeline, loops, and screens
01: ML-direction engineering manager, onsite rounds
Interview period: 2026 Q4 · Report date: November 18, 2026 · Role: Engineering manager (ML direction) · Degree listed: Master's · Channel: Not stated · Candidate's rating: Neutral experience, average difficulty · Result: Rejected
- Scheduling. A recruiter reached out in early September. The candidate rescheduled three times before the loop.
- Behavioral round. Four questions the candidate remembers, plus a fifth they forgot:
- An achievement whose impact went beyond your own team. They described a product line they had led that helped their company raise a funding round and later get acquired.
- Your manager made an unreasonable request. They had two weeks to demo a product they didn't know to a customer, so they shipped the basic functions first and improved them in later versions.
- A coworker was giving customers poor service. They began reading the whole team's customer tickets themselves and led the technical side, at the manager's request, even though they weren't the manager.
- A coworker stuck with a familiar approach. They showed a teammate a more efficient way to build a feature, went back and forth a few times, and the teammate came round.
The candidate thought this round went well.
- System design round. Design the backend for read-only lookups over about a billion IDs, while the interviewer owned the front end. Third-party systems were fine only if the candidate explained how they work inside. The candidate proposed a load balancer that sends more traffic to healthier nodes, a Redis cache and Cassandra, and spent much of the time answering how those two systems work. The interviewer asked what happens if requests are spread evenly, and pushed on data in the tens of petabytes. FastPrep's practice version is Design a Read-Only Lookup Service for a Billion Records.
Result
Rejected about three hours after the design round. The candidate felt they never reached what the interviewer was looking for.
Keep practicing
- Practice design: Design a Read-Only Lookup Service for a Billion Records (report 01).
- Behavioral questions: impact beyond your team, an unreasonable request, poor customer service and a coworker set in their ways (report 01).
- More Google questions: the Google question bank. It needs sign-in.
- Hiring insights: the Google hiring insights page.
- More firsthand reports: browse the Experience feed.
Sources: Anywhere useful interview info can be found.
Comments
0No comments yet — be the first to share what you know.