I applied online. I interviewed at 6sense in Jun 2026
Interview
The first interview round was with the Hiring Manager which revolved around projects from the resume and based on the work experience and a few managerial questions related to situation handling. The second interview round was of a Difficult level and went into depth about each of the big data tools & frameworks you have worked with. But, the worst part about that was that you could clearly feel that the interviewer had gotten all of those questions from AI, which is why the moment I asked some clarification from him because the question seemed too generic & needed constraints, he wasn't particularly able to clarify it confidently and you could see it on his face. Few of these in-depth questions were good but this latter part made the experience "Not so good" for me. The interviewer was literally asking the names & specific terms within the Flink & Spark ecosystem. It just gave me the impression that they are asking a series of AI slog hard questions with a series of expected answers and not to ask too much of clarifications, rather than testing conceptually.
Interview questions [1]
Question 1
Note: Many of these questions were specifically asked based on my work experience and the tools I have worked with. It may differ as per your experience. Explain the entire architecture of Trino. What happens when you submit a query? Explain the architecture of iceberg. What is compaction in iceberg? How do you enable the kafka events in Trino? Differences between Flink and Spark Streaming. When would you use which? Name the operators used in Flink in stream joins. If you have a spark job to process a 10GB csv file, and you have 2 spark executors with 2GB each and 2 cores, will it be able to process it successfully? If yes, are there observed issues/anomalies? If no, why? For a kafka topic where the consumer has a high lag, what will you do to reduce that lag? Will just increasing partitions by sufficient of the consumers are same? How can you delete specific events from a kafka topic? If you have a hive table, which will run faster queries on it, Hive or Trino? Do you need Trino in this usecase when you just have hive tables? If the query is a simple SELECT COUNT(*) FROM table, which will execute it faster? Let's say, you observed Trino was faster. Why would that be? What are the compute engines you can use? What is Tez? What does it mean by the term "split" printed in the Trino logs.