Data Scientist Interview Prep & Rubrics

Modelling and inference rather than platform: supervised methods, features and leakage, metric choice, experimentation and statistics, and defending a result to a stakeholder.

The loop, in order

Machine Learning24 questions · 8 topicsThe modelling half of an ML loop: supervised and unsupervised methods, features and leakage, metric choice, imbalance, deep learning foundations, forecasting and ranking.
Data & AI Engineering37 questions · 8 topicsData engineering, ML engineering and LLM/RAG interview questions covering pipelines, model evaluation, overfitting, retrieval, embeddings, vector search and production AI systems.
Python20 questions · 7 topicsPython for backend, automation, and data roles, including the runtime behaviour that surprises candidates who only know the syntax.
Databases & SQL25 questions · 8 topicsQuery-writing rounds and the storage-engine reasoning behind why a query is slow, for backend, data, and DBA roles alike.
Product Management22 questions · 7 topicsThe PM loop as it is actually run: product sense, analytical, execution, and strategy rounds, each with its own scoring rubric.
Behavioural & Culture Fit21 questions · 8 topicsThe round candidates most often under-prepare, treated as a skill with structure rather than small talk, including the rubric interviewers score against.
HR, Screening & Offer18 questions · 6 topicsThe recruiter and HR rounds that gate the technical loop and decide your compensation, where most candidates improvise and lose money.

Latest for this track

hardConceptScenario

An asyncio call times out and you handle the TimeoutError, but the background task keeps running and mutates shared state a few seconds later. What happened, and how do you make the timeout actually stop the work?

A timeout requests cancellation; it does not stop code. CancelledError is raised at the task next await point, so a task inside blocking code, shielded, in an executor thread, or swallowing the exception keeps running and commits its side effects late. Fix by re-raising CancelledError, keeping blocking calls off the loop, and using asyncio.timeout with a TaskGroup.

Concurrency In Python4 minmid, senior, staff

hardScenarioDesignConcept

A transform has been writing wrong revenue figures for three days and six downstream tables have consumed it. How do you backfill the corrected data without double-counting anything?

Make the write idempotent before you make it correct: partition-level replace rather than append, so rerunning a day produces the same result whether it runs once or five times. Then work out how far the contamination spread, because a downstream table that accumulated rather than recomputed will not fix itself when the source is repaired.

Data Pipelines4 minmid, senior, staff

mediumConceptScenarioCase Study

A churn model scores 0.94 AUC offline and barely beats the baseline in production. One of its strongest features is populated by a process that runs after the customer has already churned. Explain what happened and how you would fix it.

The feature encodes the outcome, so the model reads the answer rather than predicting it. Offline the label and the feature come from the same snapshot, which is why validation looks excellent; in production the feature is empty at scoring time and the signal vanishes. Fix by defining features by when they were available and validating on a forward time split.

Ml Fundamentals5 minmid, senior, staff

hardScenarioCase StudyBehavioural

Your largest customer, worth 15% of revenue, wants a feature nobody else has asked for, and renewal is in eight weeks. How do you decide whether to build it?

Separate the leverage from the merit: the renewal makes the request urgent, it does not make it valuable. Work out the real problem underneath the requested solution, test whether it generalises to other accounts, and price the full cost including permanent maintenance rather than the build estimate. Then choose deliberately between roadmap, configuration, services and no.

Prioritisation4 minmid, senior, lead

hardConceptScenarioDesign

A table has forty million rows, thirty-five million of them soft-deleted, and every query for active rows has got slower. Would a partial index help, and what would you have to be careful about?

A partial index only contains rows matching its predicate, so indexing WHERE deleted_at IS NULL stores five million entries instead of forty million - smaller, shallower, and far likelier to stay cached. The catch is the planner only uses it when it can prove the query predicate implies the index predicate, and it does nothing about dead rows still in the heap.

Indexing5 minmid, senior, staff

hardConceptScenarioCase Study

Your A/B test was configured for a 50/50 split but the data shows 53/47 across two million users. The result is significant and positive. What do you do?

Treat it as a broken experiment, not a rounding issue. At two million users a 53/47 split is astronomically unlikely by chance, so something removed users non-randomly - and that something almost certainly correlated with the outcome. The result cannot be trusted in either direction, so find the mechanism, fix it and rerun rather than adjusting the numbers.

Statistics And Experimentation5 minmid, senior, staff