InterviewStack.io LogoInterviewStack.io

Debugging and Recovery Under Pressure Questions

Covers systematic approaches to finding and fixing bugs during time pressured situations such as interviews, plus techniques for verifying correctness and recovering gracefully when an initial approach fails. Topics include reproducing the failure, isolating the minimal failing case, stepping through logic mentally or with print statements, and using binary search or divide and conquer to narrow the fault. Emphasize careful assumption checking, invariant validation, and common error classes such as off by one, null or boundary conditions, integer overflow, and index errors. Verification practices include creating and running representative test cases: normal inputs, edge cases, empty and single element inputs, duplicates, boundary values, large inputs, and randomized or stress tests when feasible. Time management and recovery strategies are covered: prioritize the smallest fix that restores correctness, preserve working state, revert to a simpler correct solution if necessary, communicate reasoning aloud, avoid blind or random edits, and demonstrate calm, structured troubleshooting rather than panic. The goal is to show rigorous debugging methodology, build trust in the final solution through targeted verification, and display resilience and recovery strategy under interview pressure.

HardTechnical
90 practiced
Given a sharded logging schema: inference_logs(shard_id, request_id, model_version, prediction, timestamp), write a SQL query that finds request_ids where model_version A and B produced different predictions within 1 second of each other. Consider cross-shard joins and efficient patterns for large log tables.
MediumTechnical
69 practiced
Explain how you would use property-based testing (e.g., Hypothesis) to discover edge cases in an NLP tokenization and embedding pipeline. Provide concrete examples of properties to test, the kinds of random inputs to generate (unicode, control chars, long sequences), and how to triage failures found by the fuzzer.
MediumTechnical
87 practiced
Describe and sketch code for a checksum-based dataset integrity system for an ETL pipeline: compute per-row and per-file checksums, store metadata, detect schema drift, and allow partial repairs (recompute affected files). Discuss handling schema evolution and performance for large datasets.
EasyTechnical
71 practiced
Describe how you would perform randomized stress testing for a preprocessing merge step that joins two tables (features and labels) under time constraints. Include types of randomization (missing keys, duplicate keys, type mismatches), how to generate test cases, and how to make the tests fast but representative.
MediumTechnical
134 practiced
Write a pytest test suite for a feature-encoding function that must: handle empty inputs, map unseen categories to an OOV token at inference, and preserve deterministic ordering when a seed is provided. Show test cases and assertions (pseudocode acceptable). Explain how to mock external mapping files in CI.

Unlock Full Question Bank

Get access to hundreds of Debugging and Recovery Under Pressure interview questions and detailed answers.

Sign in to Continue

Join thousands of developers preparing for their dream job.