Back to posts 
Interview experience
Google · Software Engineer · Mid-Level
Community-sourced A factual summary of an interview experience shared publicly outside interviewd.
Nov 1
In Progress
Overall Experience
A two-round process for a GenAI software engineering role: a data structures round on grid connectivity problems with a follow-up, and a requirements and design round for an AI agent in manufacturing. The design round included follow-up questions on hallucination prevention and permissions. The candidate reported running short of time in the first round and answered the design round normally.
2 rounds
I
Posted by interviewdInterview Process (2 rounds)
Round 1 of 2: Coding / DSA
60 min
Question 1
MediumNumber of Islands: given an m x n grid of '0' (water) and '1' (land), return the number of islands. Land cells connect horizontally or vertically, and cells outside the grid count as water. Input is m and n, followed by m binary strings of length n. Constraints: 1 <= m, n <= 300 and m * n <= 90,000.
LeetCode equivalentExperience
The candidate started with this problem and completed it.
Question 2
MediumNumber of Islands II: an initially all-water m x n grid receives k operations, each turning one position (r, c) into land. After each operation, output the current island count; a position that is already land leaves the count unchanged. Input is m, n, and k, followed by k lines of r c. Output the k counts separated by spaces. Constraints: 1 <= m, n <= 10,000 and 1 <= k <= 100,000. The whole grid should not be rescanned after each operation.
LeetCode equivalentExperience
This was the dynamic island-addition follow-up. The candidate did not finish it because of time management.
60 min
Question 1
MediumDesign an AI agent for manufacturing customers that supports three capabilities: looking up production progress (work-order status, current operation, planned versus actual output, estimated completion time, and delay reasons); creating work orders, with validation of materials, capacity, routing, and required fields; and generating customer quotations from material, process, quantity, lead time, and contract terms. Discuss the architecture and interactions among the agent, the LLM, tools, and manufacturing systems such as MES, ERP, and quoting services; correctness, auditability, and idempotency for write operations; reducing hallucinations for production status, inventory, prices, and delivery dates; authentication, authorization, and tenant and data isolation; and failure handling, retries, human approval, and audit logs.
Experience
Follow-up questions asked how to prevent hallucinations and how to implement permissions. The candidate's approach and outcome were not reported.
Loading comments…