Data Engineer Interview Questions for India 2026

SQL remains the core language of data engineering: tested at every level. Beyond SELECT and JOIN, interviewers test: window functions (ROWNUMBER, RANK, DENSERANK, LAG, LEAD, PARTITION BY): these appear in almost every senior data engineering interview. CTEs (Common Table Expressions): recursive CTEs for hierarchical data. Performance optimisation: explain plans, index strategies, query rewriting. Data modelling: star schema (fact tables + dimension tables) vs snowflake schema (normalised dimensions). SCD (Slowly Changing Dimensions) Type 1 (overwrite), Type 2 (new row), Type 3 (previous value column). Partitioning and bucketing in distributed storage.

Spark is the dominant distributed processing framework tested at Indian product companies. Core concepts: RDD vs DataFrame vs Dataset. Lazy evaluation: transformations (map, filter, groupBy) are lazy; actions (count, collect, write) trigger computation. Partitioning: repartition (shuffles data), coalesce (reduces partitions without full shuffle). Caching: persist() and cache(): when and why. Broadcast joins: for small tables (< a few hundred MB) to avoid shuffle. Common interview scenarios: handling data skew (salting technique), optimising Spark jobs that run slowly, reading from and writing to Parquet, Hive, and Delta Lake.

Airflow is the most common orchestration tool tested in Indian data engineering interviews. Key concepts: DAG (Directed Acyclic Graph): the core abstraction, collection of tasks with dependencies. Operators: PythonOperator, BashOperator, SparkSubmitOperator, BigQueryOperator, S3Operator. Task dependencies: >> and << operators. XComs: passing small data between tasks. Sensors: waiting for external conditions (file arrival, API response). Scheduling: cron expressions in Airflow. Common design questions: how to handle late-arriving data, backfill strategies, idempotent pipelines (safe to re-run), circuit breakers for upstream failures.

Modern data engineering at Indian companies typically uses cloud data stacks. AWS: S3 (storage), Glue (ETL), Redshift (data warehouse), Kinesis (streaming), Athena (serverless SQL). GCP: GCS (storage), Dataflow (Apache Beam), BigQuery (serverless data warehouse), Pub/Sub (streaming). Azure: ADLS (storage), ADF (pipeline orchestration), Synapse Analytics (data warehouse). dbt (data build tool): SQL-first transformation layer, increasingly tested. Growing in Indian companies as the standard for transform layer. Delta Lake / Apache Iceberg: ACID transactions on data lakes: enables upserts and time travel on large datasets. Expected knowledge at senior levels.

Data engineering has become one of the highest-paying technical specializations in India. Entry-level data engineers with 1-2 years earn 8-14 LPA, typically at analytics consulting firms or mid-size startups. Mid-level engineers with 3-6 years and strong Spark and Airflow skills earn 20-38 LPA at companies like Flipkart, Swiggy, Juspay, or Meesho. Senior data engineers and principal engineers at unicorns or FAANG India offices earn 45-80 LPA. Candidates with expertise in dbt for transformation layer design, Delta Lake or Apache Iceberg for lakehouse architectures, and real-time streaming on Flink or Kafka Streams command the top of the range and face strong competition from returning NRI professionals.

Real-time data pipeline design is a core interview topic at Swiggy, Zepto, Flipkart Quick Commerce, and ride-sharing platforms like Rapido. Expect questions on Kafka consumer group rebalancing and its impact on end-to-end latency, exactly-once semantics versus at-least-once delivery guarantees, and how to handle schema evolution without breaking downstream consumers using Schema Registry. Interviewers at Swiggy have asked candidates to design a system that streams order state changes with sub-second latency to delivery partner apps. Practice designing fault-tolerant pipelines that handle backpressure, partition skew in Kafka topics, and late-arriving events using watermarking in Spark Structured Streaming or Apache Flink.

Indian product companies have matured significantly in data quality practices and now test candidates on data reliability engineering. Interviewers at Razorpay, Groww, and Zerodha ask about implementing data quality checks using Great Expectations or custom Spark assertions, and about handling schema drift in incoming data from third-party APIs. Data observability concepts borrowed from tools like Monte Carlo and Bigeye are increasingly discussed even if companies build similar checks in-house. Expect to explain how you would detect a sudden spike in null rates or anomalous row counts in a production pipeline without manual intervention. Building lineage metadata and tracking freshness SLAs for business-critical tables is another frequently discussed engineering challenge.