Existing Codebase Debugging Round

Debug an unfamiliar existing codebase under time pressure. The task is less about inventing a new algorithm and more about navigating files, running tests, identifying a few small defects, and communicating the debugging path. Requirements Environment: IDE-like workspace or existing multi-file codebase. Candidate can run the existing code and tests. Fix several seeded bugs; one MLE version describes 3 or 4 bugs, each requiring only 1-2 lines once identified. SWE loop variant centers on an "assign project" codebase and is time-constrained. Example bug shapes: A loop forgets to call a reset function. Filtering uses the wrong identifier field, such as filtering by id instead of course_id. Communicate while navigating the codebase. The interviewer may provide help, but the round values how quickly the candidate builds a mental model of unfamiliar files. Notes The bugs can be easy after discovery; the hard part is locating them under time pressure. Start by running tests or the failing scenario, reading the stack trace / assertion, then tracing data flow to the smallest responsible function. Avoid broad rewrites. Patch the narrow defect, rerun the relevant test, then move on. Keep a running hypothesis list out loud: what failed, where the data changes, what invariant should hold, and how the patch restores it. Preparation Practice debugging a 3-5 file Python project with tests already present; timebox yourself to 45 minutes. Drill common defect classes: wrong id field, stale state not reset, off-by-one loops, filter predicates, missing validation, and mutation of shared state. Rehearse a concise narration style: observation, hypothesis, experiment, patch, verification.