DSA for Data Engineers: Focused 100–200 Pattern Roadmap

Data Engineering interviews usually test DSA at an easy-to-medium level, with a bias toward problems that look like data processing. You do not need hundreds of hard puzzles. You need the recurring patterns, practised until they are automatic.

How to practise

1. Learn the pattern, then solve problems in order of difficulty.
2. State the brute force first, then optimise, then give time and space complexity.
3. Write clean Python: clear names, small helper functions, edge cases (empty input, duplicates, ties).
4. Re-solve problems you got wrong a week later.

The ranges per stage add up to roughly 100 to 200 problems in total. Stop a stage early once problems feel routine.

The plan

1. 01

   Stage 1: Arrays, strings and hashing

   Frequency counts, grouping, deduplication and lookups with dicts and sets.

   Typical effort

   20–30 problems

   Outcome

   You default to a hash map when you need lookups and can state the complexity.

   • Python Data Structures for Data Engineering Interviews
   • List vs tuple vs set: when would you use each in a data pipeline?
2. 02

   Stage 2: Two pointers and sliding windows

   Pairs in sorted data, longest substring or subarray with a constraint, moving aggregates.

   Typical effort

   20–30 problems

   Outcome

   You recognise window problems and keep them O(n).

3. 03

   Stage 3: Sorting, intervals and merging

   Merge overlapping intervals, meeting rooms, merging sorted streams.

   Typical effort

   15–25 problems

   Outcome

   You can reason about interval overlap, which also appears in sessionisation and SCD logic.

   • Slowly Changing Dimensions: Types 0 to 6
4. 04

   Stage 4: Heaps and top-K

   Top K frequent items, K-way merge, running medians.

   Typical effort

   10–20 problems

   Outcome

   You use a heap for top-K and streaming problems.

   • Python Iterators, Generators and Memory-Efficient Processing
5. 05

   Stage 5: Stacks and queues

   Matching brackets, monotonic stacks, BFS with deques.

   Typical effort

   10–15 problems

   Outcome

   You choose a deque for queues and know why list.pop(0) is slow.

6. 06

   Stage 6: Binary search

   Search in sorted data and on the answer space.

   Typical effort

   10–15 problems

   Outcome

   You write bug-free boundary conditions.

7. 07

   Stage 7: Graphs and trees (basics)

   BFS, DFS, topological sort for dependencies (the same idea behind DAG schedulers).

   Typical effort

   15–25 problems

   Outcome

   You can order tasks with dependencies and detect cycles.

   • Airflow DAGs, Scheduling, Retries and Task Dependencies
8. 08

   Stage 8: Data-processing problems

   Parse logs, aggregate records, deduplicate events, join two lists, in Python and SQL.

   Typical effort

   15–30 problems

   Outcome

   You solve 'process these records' problems cleanly in both languages.

   • How would you detect and remove duplicate records safely?
   • What is a generator and why can it help with large datasets?