Did the Intel data engineering loop in Q1 this year. Took the role. Posting because when I was prepping I couldn't find a single recent, detailed account of what they actually ask.
Intel has a decent-sized data engineering function spread across teams: manufacturing analytics, product telemetry, internal tooling for chip development metrics, and the AI/HPC side. My loop was for the manufacturing analytics side, which involves ETL pipelines moving data from fab floors and sensor arrays into analytical systems.
SQL: Expect this to be significant. I had two dedicated SQL exercises. One was a fairly advanced aggregation problem with window functions, partitioning, and some nasty NULLs scattered throughout. The second was an optimization problem: here's a slow query, here's the schema, what would you change? I had to explain index strategy and denormalization tradeoffs. This is not entry-level SQL. If your SQL has gaps in window functions or query plans, fix that before your loop.
Pipeline design: One round was effectively a system design for pipelines. Design a data ingestion system that can handle streaming telemetry from thousands of chip test benches, with variable schemas and occasional sensor failures creating incomplete records. They cared about schema evolution (how do you handle a sensor that starts sending a new field?), exactly-once processing guarantees, and data quality checks. I talked through Apache Kafka for ingestion, Spark for transformation, Delta Lake for storage. They were familiar with all of it and asked specific follow-ups about Delta's ACID properties.
Python / Pyspark coding: One session, medium difficulty, focused on data manipulation. Write a function to do X transformation on this DataFrame. Nothing algorithmic in the LeetCode sense, but I had to write clean, efficient Spark code.
Behavioral: Standard stuff, but framed around cross-team data ownership questions. They really care about how you handle situations where engineering teams don't want to change their schema or instrumentation. That's a real problem at Intel.
Total loop was 5 weeks, received offer in 10 days after final round.