Datetime data in pandas https://uvik.io/ is handled through the datetime64 type and the .dt accessor, which gives you access to components like year, month, and day of the week. This is a test of whether you understand NumPy at a conceptual level, not just as a syntax library. At 50 million rows, this kind of choice can have a major impact on runtime.
Start by solving concrete problems and validating results on realistic data. Do these drills and say these lines aloud to show production thinking in interviews that ask PySpark data engineer interview questions, too. It is designed for data engineers who already know Python and SQL but need structured, time-bound preparation to convert that knowledge into interview-ready execution. Make jobs idempotent, add checkpoints for streaming, tune executor memory and cores, persist hot datasets with caching, and include metrics and alerts. In answers, emphasize writing clear, Pythonic code and consider scalability, e.g., using generators or batch processing to handle big data. A 45–60 minute live call where you’ll write SQL queries, discuss schema design, and possibly code in Python.
Translates the logical model into a representation that considers the implementation details. Data modeling is a structured approach to designing a data storage system, whether it’s a database, data warehouse, or any other data repository. These 100 questions provide a strong starting point for interview readiness. With thorough study and practice, candidates can excel in Python data science interviews. Candidates should be ready to write code during interviews.
Intermediate Level Questions
- Against processes you win on startup cost and memory; against threads you win on isolation.
- A DataFrame is a 2-dimensional labeled data structure with columns of potentially different types.
- Since data scientists come from a variety of backgrounds, including software development or more pure statistics, the level of coding ability expected in an interview will differ.
- The second is to use nlargest() or nsmallest(), which skips the full sort and goes straight to the rows you care about.
- Create a dictionary with list elements as keys and their occurrences as values.
- You should list all the tools in which you have expertise and pick one as your favourite.
The follow-up they askThe log has duplicate timestamps for one entity. The follow-up they askRedefine ‘consecutive’ as business days only. The follow-up they askMarketing wants sessions to also break on UTM source change. This ‘ratio to total’ shape generalizes to percent-of-day, percent-of-cohort, and contribution analysis.
More Interview Prep Guides
The syntax is small and readable, so the first hundred lines come quickly. That’s why a lot of what follows is some version of “when would you not do this” — there’s no memorized answer to those. This is the list of Python questions I use when interviewing senior developers. Join 500+ engineers already practising on DataCodingHub. “Loved the real code execution — no more guessing if my solution is right. The instant pass/fail made my preparation so much more focused and efficient.” “Finally a platform that understands what data engineers actually do. No more irrelevant LeetCode grinds. Highly recommend.”
The distributed-compute checks that survive in 2026 loopsWork through 3 problems → The follow-up they askYour sink is an external API with no idempotency support. The follow-up they askYour watermark is too aggressive and 2% of conversions land late. The follow-up they askProduct wants exactly-once into the OLAP store. The follow-up they askThe source has no updated_at and you cannot install CDC.
Can you explain the difference between Python’s deep and shallow copy?
Trips-and-payments modeling and SCDs come up, and so do streaming consumers. Take-home plus debugging focus, exactly-once and reconciliation framing. Hiring committees average across rounds, so consistent strength everywhere beats one spectacular hour with a weak one attached; a single bad round drags the whole packet at committee-driven companies.
When a thread performs blocking operations such as reading from a file, making a network request, or waiting for a database response, the GIL is released. Even if you create several threads, they take turns executing, which limits performance gains for tasks like mathematical computations, data processing, or heavy algorithms. Its primary purpose is to simplify memory management and maintain thread safety for Python’s internal data structures. You can’t build this on a streaming-first architecture and reliably promise 8am delivery.