pandas interviews

Work in the tool while you explain your decisions.

Choose a role and level. Each interview has a new task, follow-up questions and a report that checks your work and your explanation.

Take the assessment Choose an interview

Speak or type. Sign in to save your results across devices.

Timed assessment

12 practical questions
50 minutes
Mid level

Questions are drawn from a bank of 36. Your report includes a skill breakdown and lessons to practice next.

Start assessment

Practice interviews

Junior tasks cover the fundamentals. Mid and senior tasks add debugging, changing requirements and review decisions.

Position

10 interviews

Data analyst
Junior
20 min

Support backlog with unknown owners

At Mossline Support, count open tickets by owner and sum their effort into columns owner, tickets and effort. Include missing owners as Unassigned. Only status open belongs in this backlog. A zero effort is a recorded value.

Start interview

Skills and rubricMissing groups and named aggregation

Filter the open population first. Preserve missing owners, then count rows and sum the recorded effort.
Identifies a source row that exposes the named mistake.
Explains the result using the supplied records and a concrete test.

Data analyst
Junior
20 min

Supplier lead-time quarantine

Cedar Parts received a dirty supplier export. Parse days as numeric, exclude unparseable or negative lead times, and return supplier and mean_days for suppliers with at least one valid observation. Preserve measured zero days. Store the result in answer.

Start interview

Skills and rubricNumeric parsing with an audit trail

Coerce invalid text to missing, then select nonnegative measurements. A missing value and a recorded zero have different meanings.
Identifies a source row that exposes the named mistake.
Explains the result using the supplied records and a concrete test.

Data engineer
Junior
20 min

Shipment enrichment without invented matches

Enrich every row of shipments with region from accounts. Return shipment and region. An absent or missing account key has region Unknown. A missing key in accounts is not a matching business account. Keep the shipment population.

Start interview

Skills and rubricLeft joins and missing business keys

Retain the left population and validate the right key grain. Missing reference keys cannot identify an account.
Identifies a source row that exposes the named mistake.
Explains the result using the supplied records and a concrete test.

Data engineer
Mid
30 min

Debug the newest parcel status

Parcel status snapshots arrived out of order. Return one row per parcel with parcel and status: choose the greatest updated timestamp, then greatest revision for a timestamp tie. The supplied answer keeps the last physical row and is wrong. Repair it.

Start interview

Skills and rubricDeterministic snapshot deduplication

Select by event ordering, then project the requested status. State the rule for an exact duplicate revision before processing a production refresh.
Defends the revised contract and tests its exact boundary without hard-coded output.
Reconciles the output against an independently chosen boundary example.

Product analyst
Mid
30 min

A product day across UTC midnight

A product manager wants daily distinct visitors from page events. Inspect timestamps and describe the day boundary you would choose before building the report.

Start interview

Skills and rubricLocal-day activity and changed requirements

Parse timestamps as UTC instants, convert to the business zone and only then extract the date. Distinct visitors are counted within each local date.
Defends the revised contract and tests its exact boundary without hard-coded output.
Reconciles the output against an independently chosen boundary example.

Data analyst
Mid
30 min

Review an uneven service rate

A reviewed report averages daily completion rates, giving a two-job day the same weight as a twenty-job day. Repair it: return team and rate as total completed divided by total offered. Use missing for a team with zero offered. Don't invent extra offers. Explain how you would remove row loops before measuring performance.

Start interview

Skills and rubricWeighted metrics and vectorized aggregation

Aggregate the numerator and denominator at the requested grain. Divide after aggregation and keep an explicit zero-denominator policy.
Defends the revised contract and tests its exact boundary without hard-coded output.
Reconciles the output against an independently chosen boundary example.

Data scientist
Mid
30 min

A feature that cannot see its own target

Build a leakage-safe prior_units feature for each account observation. Sort each account by at, then observation for ties. prior_units is the sum of units from earlier rows of that account, excluding the current observation. The first observation gets zero. Return observation and prior_units.

Start interview

Skills and rubricPrior-only group features

Sort at the entity grain and remove the current measurement from the cumulative total. An arbitrary physical order leaks or changes information.
Defends the revised contract and tests its exact boundary without hard-coded output.
Reconciles the output against an independently chosen boundary example.

Data engineer
Senior
40 min

Telemetry known when an alert fired

For every alert, attach the most recent reading from the same device at or before the alert timestamp, at most five minutes old. Keep unmatched alerts with missing value. Return alert and value. A future reading must never explain an earlier alert.

Start interview

Skills and rubricBackward temporal enrichment

Use a backward match within each device and a bounded tolerance. Sort by the temporal key before merge_asof.
Explains silent failure modes, ownership and a measurable scale plan.
Defines a measurable runtime and memory check without unsupported speed claims.

Product analyst
Senior
40 min

Review an activation denominator

Review a draft activation report that counts event rows as activated accounts. Return cohort, accounts and activated. cohort is the YYYY-MM signup month. An account is activated if it has any activate event within seven days of signup, including the signup instant but excluding exactly day seven. Count each account once and retain cohorts with no activation.

Start interview

Skills and rubricDistinct populations and complete denominators

Build the denominator from signups, use event timestamps to mark membership, and aggregate accounts once. Don't infer activation from the number of events.
Explains silent failure modes, ownership and a measurable scale plan.
Defines a measurable runtime and memory check without unsupported speed claims.

Data engineer
Senior
40 min

Session counts after replayed events

Count sessions per visitor from the Elm Help event stream. Deduplicate event_id first, then order each visitor by timestamp. A new session begins on the first event or after a gap strictly greater than thirty minutes. A gap of exactly thirty minutes stays in the same session. Return visitor and sessions.

Start interview

Skills and rubricDeduplicated event sessionization

Deduplicate at event grain, compare adjacent events inside each visitor and count session starts. The threshold equality is part of the metric contract.
Explains silent failure modes, ownership and a measurable scale plan.
Defines a measurable runtime and memory check without unsupported speed claims.

Choose your character

pandas interviews

Timed assessment

Practice interviews

Support backlog with unknown owners

Supplier lead-time quarantine

Shipment enrichment without invented matches

Debug the newest parcel status

A product day across UTC midnight

Review an uneven service rate

A feature that cannot see its own target

Telemetry known when an alert fired

Review an activation denominator

Session counts after replayed events

Interview Hub access

Your Work Gym account