Klarna India Data Engineer Interview: Rounds, Questions and Preparation

What the job involves

Data Engineers build and maintain the pipelines that move and transform data from source systems into warehouses and lakes for analytics and machine learning. In a global fintech like Klarna India, you would work on batch and streaming pipelines, ensure data quality and governance, and support real-time decisioning at scale. Day to day involves writing SQL and Python, orchestrating jobs, and collaborating with analysts and product teams.

Role and technical knowledge

1. 1

   Explain the difference between ETL and ELT, and when you would choose one over the other in a fintech data pipeline.

   Cover transformation before vs after loading, and mention cloud warehouses like BigQuery or Snowflake that make ELT practical for large volumes and flexible schemas.

2. 2

   How would you design a star schema for a payments transaction dataset?

   Describe fact table (transactions) with measures like amount, and dimension tables for customer, merchant, date, and payment method. Mention surrogate keys and slowly changing dimensions.

3. 3

   What are the key differences between batch and stream processing, and give a fintech use case for each.

   Batch for daily settlement reports and reconciliation; stream for real-time fraud detection or transaction authorisation. Mention latency, throughput, and tools like Spark and Kafka.

4. 4

   How do you ensure data quality in a pipeline?

   Mention validation checks (nulls, duplicates, schema drift), automated tests, alerting, and data contracts. Give an example like checking that transaction amounts are positive.

Practical scenarios and problem solving

1. 5

   A daily pipeline that loads transaction data into the warehouse failed last night. How do you debug and recover?

   Check logs and error messages, verify source availability, validate schema changes, rerun failed tasks idempotently, and backfill missing partitions. Mention communication with stakeholders.

2. 6

   You need to join a large transactions table with a customer table, but the query is slow. How would you optimise it?

   Discuss partitioning and clustering, filtering early, using broadcast joins for small tables, avoiding SELECT *, and checking execution plans. Mention indexing or sort keys.

3. 7

   How would you handle late-arriving data in a streaming pipeline for fraud alerts?

   Use event-time processing with watermarks, allow a grace period, and update results in a stateful store. Mention trade-offs between latency and completeness.

4. 8

   You notice duplicate records in a reporting table. How do you investigate and fix the root cause?

   Trace back through the pipeline, check join keys and deduplication logic, review idempotency of writes, and add a unique constraint or dedup step. Suggest monitoring to prevent recurrence.

Behavioural and HR questions

1. 9

   Tell me about a time you had to learn a new technology quickly for a project.

   Use STAR format: situation, task, action, result. Highlight self-learning resources, a concrete outcome, and how you applied it to deliver value.

2. 10

   Describe a situation where you disagreed with a teammate on a technical approach. How did you handle it?

   Show respect, data-driven discussion, and willingness to compromise. End with a positive outcome and what you learned about collaboration.

3. 11

   Why do you want to work as a data engineer in a global fintech company?

   Connect personal interest in large-scale data and financial systems, mention the impact of reliable data on customer experience, and show awareness of the company's domain.

4. 12

   How do you prioritise when you have multiple data requests from different teams?

   Talk about understanding business impact, urgency, and effort. Mention communicating timelines, using a ticketing system, and escalating when needed.

Coding and SQL deep dive

1. 13

   Write a SQL query to find the top 3 merchants by total transaction amount for each month in the last year.

   Use a window function like RANK() or DENSE_RANK() partitioned by month and ordered by sum(amount) descending, then filter rank <= 3. Mention handling ties and date filtering.

2. 14

   Given a list of transactions with timestamps, write a Python function to detect customers with more than 5 transactions in any 10-minute window.

   Use a sliding window with two pointers or a deque after sorting by customer and time. Discuss time complexity and edge cases like multiple customers.

3. 15

   How would you find and remove duplicate rows in a table using SQL?

   Use ROW_NUMBER() over a partition of all columns or key columns, then delete rows where row number > 1. Mention using a CTE and testing on a backup.

4. 16

   Explain how you would implement a slowly changing dimension type 2 in a data warehouse.

   Describe adding effective date, end date, and current flag columns. Use merge or upsert logic to expire old records and insert new versions when tracked attributes change.