Interview guides · Kodiak AI

Kodiak AI Data Scientist Interview Questions & Guide 2026

Take-home style practice for data scientist candidates: a messy dataset, a fixed time budget, and an analysis you write up and defend.

The 60-second brief

As a Data Scientist at Kodiak AI, you play a pivotal role in advancing autonomous vehicle technology through data-driven insights and innovative solutions. This position is essential for enhancing the safety, efficiency, and performance of autonomous systems. You will work closely with advanced algorithms, machine learning models, and vast datasets to solve complex challenges that arise in real-world environments, such as traffic patterns, telemetry data, and vehicle positioning.

Coding rounds for this role are usually data-manipulation shaped rather than data-structure shaped: group-bys, joins, time windows, ranking within a partition. Confirm the format before spending a week on graph traversal.

Focus areas

Strip CI, retry and synthetic traffic first
Separate contracted seats from actively used seats
Power experiments for heavy-tailed account revenue

Overview

Your expertise will directly impact the development of products that not only improve the driving experience but also contribute to a safer transportation ecosystem. By collaborating with multidisciplinary teams, including engineering, product management, and AI research, you will influence strategic decisions and drive the progression of Kodiak AI’s initiatives. Expect to engage with cutting-edge technologies and tackle intricate problems that require deep analytical thinking and a strong technical foundation.

The Kodiak AI interview process

Kodiak AI candidates report 4 rounds · ≈ 3-5 weeks. The stages below are what candidates describe, not a published process.

Phone Screening

Data Scientist covers at least four different jobs: experimentation, product analytics, causal work on observational data, and applied modelling that ships into a system. A screening call is the cheapest place to find out which of them is being hired for, and doing that diagnosis openly reads as senior rather than fussy. Ask what the last few pieces of work on the team actually were, and roughly how a week splits between querying, modelling and stakeholder time. Then say which parts of that you have done and which you have not. Claiming the whole range is the fastest way to be caught one round later.

What to demonstrate

Whether you can distinguish the flavours of the role and locate your own experience inside one of them honestly
Whether you name what you have not done instead of stretching to cover every line of the posting
Whether your hard constraints (notice period, location, work authorisation, level) surface now rather than at offer stage

How to prepare

Map the last two years of your time into rough percentages across query writing, experiment design, modelling and stakeholder work, so a question about scope has a real answer
Mark every responsibility in the posting as done, adjacent or new, and prepare one sentence for each adjacent item naming the closest thing you have actually built
Decide which logistics are non-negotiable before the call so you can state them in one sentence rather than negotiating live

Technical Interviews

A handful of shapes account for most of what gets asked in this format: a ranking or deduplication inside groups, a running or rolling total, a period-over-period comparison, and a cohort tracked forward over time. Recognising the shape quickly is most of the speed here; deriving it from scratch while a clock runs is where the time goes. Know that a window function keeps every row while a GROUP BY collapses them, and know which one the question needs. If the exercise is in Python instead of SQL, the same shapes arrive as groupby with transform, shift and merge, and the same grain mistakes are available.

What to demonstrate

Whether you reach the right construct without a detour, such as ROW_NUMBER over a partition to deduplicate instead of a self-join against a MAX subquery
Whether you know what your window frame actually is, since adding ORDER BY inside OVER changes the default frame and silently changes a running total
Whether the thing runs. A near-miss that throws an error scores below a plainer query that returns the right rows.

How to prepare

Write each of the four shapes once from memory against a small schema and keep the working version somewhere you will reread it: dedupe with ROW_NUMBER, a running total, a month-over-month change with LAG, and a retention table
Compute one running total twice on data with tied timestamps, once on the default frame and once with ROWS BETWEEN UNBOUNDED PRECEDING AND CURRENT ROW, and look at where the two disagree
If Python is on the table, rebuild the dedupe and the running total with groupby and cumsum, then assert the two implementations return identical rows

Behavioral Interviews

This round decides whether you owned a decision or watched one happen nearby. Interviewers for data roles listen for the point where the analysis stopped being a report and started changing what someone did, so build each story around that hinge: what was going to happen by default, what you found, and what happened instead. The most common weakness is a story that ends at delivery. If you can name the decision your work changed and the number that moved because of it, most follow-ups become easy.

What to demonstrate

Whether the decision was yours to influence, or whether you are narrating a team outcome in the first person
The counterfactual: what would have been done without your analysis, and why that default was worse
How far your involvement ran past the handoff, and whether you checked that the change did what you predicted

How to prepare

Pick three projects and write one sentence for each naming the decision-maker, the choice in front of them, and what they chose after seeing your work. If you cannot name a person and a choice, the story is not ready for this round.
Reconstruct the baseline for your strongest project from the original query or dashboard rather than memory, so the before-number survives a follow-up asking where it came from.
Prepare an honest version of a project where your recommendation was overruled, including what you did with the analysis afterwards.

Final Interviews

Where a loop includes a partner from outside the data team, that conversation usually carries the same weight as the technical ones and gets the least preparation. The person opposite you will not follow a derivation and does not need to. They are working out whether having you involved would make their decisions better or slower. The failure mode is not being too technical. It is answering a question about a decision with a description of your method, leaving the translation to them. What they carry into the debrief is the sentence you handed them, not the analysis underneath it.

What to demonstrate

Whether a statistical result arrives as something the partner could act on, with the one caveat that would change their decision kept and the rest left out
Whether you can state what you need from their side, in their terms: instrumentation that does not exist yet, a definition they own, or a holdout they have to agree to
Whether uncertainty is given as a range someone can plan against, rather than as hedging that invites them to ignore the result
Whether you ask what decision is actually on the table before explaining anything

How to prepare

Take a result you know well and write the version for someone who stops reading after one sentence, then the three-minute version, and check the short one is not the long one with the qualifications stripped out
For a past project, list everything you asked a non-technical partner for and how you phrased it, then rewrite each ask so it names what goes unmeasured without it
Practise saying where a result does not apply, out loud, in one sentence that a partner could repeat accurately to someone else

Where candidates lose points

Reporting a mean over accounts when account revenue is heavy-tailed

When a small number of accounts hold most of the revenue, the sample mean is dominated by whichever of them happens to be in the sample, and the sample variance keeps growing as more data arrives instead of stabilising. In that regime the usual central-limit-based confidence interval understates uncertainty, and a single renewal or a single large account's batch job can flip the sign of a measured effect. The fixes are to pre-register a winsorisation or capping rule before looking at the outcome, to report account counts crossing a threshold alongside the revenue figure, or to define the estimand on a bounded transform. Choosing the cap after seeing the result is a separate and worse problem, because the cap then encodes the answer.

Randomising an experiment at the user level when users share an account

Two problems fire at once. Colleagues in one workspace see each other's work and talk to each other, so a treated user changes the behaviour of a control user in the same account, which violates the no-interference assumption and biases the estimate toward zero. Separately, outcomes within an account are strongly correlated, so the effective sample size is roughly n / (1 + (m - 1) * rho) for m users per account and intra-class correlation rho, not n. With rho around 0.3 and twenty users per account that is a design effect near 6.7, meaning a user-level confidence interval is about two and a half times narrower than it should be and results cross significance thresholds on noise alone. Randomise the account and cluster the standard errors.

Averaging per-user rates to produce a population rate

Decide which quantity you want: the mean of per-user ratios and the ratio of summed numerator to summed denominator are different estimands, and heavy users dominate one but not the other. For a ratio metric, aggregate numerator and denominator separately and use the delta method for its variance.

Extrapolating a first-week lift inflated by novelty effects

Plot the treatment effect by days since first exposure instead of quoting one pooled average. A lift that decays toward zero across the test window is behaviour that will not persist, and annualising it produces a forecast that misses by an order of magnitude.

Questions & practice

Given a dataset, how would you implement a basic linear regression model in Python?

Medium · machine learning and modelling

Approach

Set a baseline first, so any model has something honest to beat.
Check what information would not exist at prediction time, and exclude it.
Frame the prediction: the label, the moment of prediction, and the action it triggers.

Follow-up

Where could label leakage enter this setup?
What would you monitor after launch to know the model is still valid?

Describe your process for conducting a root cause analysis on a performance issue in an AI model.

Medium · machine learning and modelling

Approach

Set a baseline first, so any model has something honest to beat.
Say how the offline result would be validated online before it is trusted.
Check what information would not exist at prediction time, and exclude it.

Follow-up

What would you monitor after launch to know the model is still valid?
How would you choose the decision threshold, and who owns that choice?

Permutation-test a consumption experiment randomised at account level

Hard · permutation-test · heavy-tails · cuped · experiment-readout

An experiment randomised 900 accounts into two arms. You have one row per account: account_id, arm, consumption_28d (billable units after launch) and consumption_pre (the 28 days before). Consumption is heavy-tailed and the largest account is several percent of the total. Write a permutation test from scratch: winsorise at the pooled 99th percentile as a pre-registered rule, use the difference in arm means of the winsorised outcome as the statistic, and obtain a two-sided p-value from 20,000 relabellings of the account-level arm vector. Report the observed effect, the p-value, and the same test on a CUPED-adjusted outcome.

Approach

Be precise about what the permutation test needs. Under the sharp null of no effect for any account, the outcomes are exchangeable across arm labels, and the test is valid for ANY statistic T(outcomes, labels) provided the identical function is applied to the observed labels and to all 20,000 relabellings. The pooled 99th percentile is a function of the outcome vector alone, so recomputing it inside the loop returns the same number 20,000 times: that is wasted CPU, not a bias, and hoisting it out is an optimisation rather than a correctness fix. Say plainly that capping at all changes the estimand from mean consumption to mean capped consumption; it is not a neutral cleaning step.
The mistake that does invalidate the test is an asymmetry between the observed statistic and the permuted ones, and the easiest way to create it is to derive the cleaning rule from the observed arm labels and then freeze it — winsorise each arm at its own observed 99th percentile, hold those two caps fixed, and permute. The observed value is then computed with caps matched to its own partition while every relabelling is scored with caps belonging to a different one, so the null distribution no longer answers the question the p-value claims to answer. A per-arm cap recomputed consistently inside every permutation is a valid test, but it estimates a contrast whose two sides are capped at different thresholds, so prefer the pooled cap on estimand grounds and pre-register it.
Permute the account-level arm vector, because the account is the randomisation unit. Relabelling anything finer — users, workspaces, requests — generates a null distribution narrower than the design actually supports and returns p-values that are anti-conservative.
Vectorise the null: tile the treatment indicator into a (B, n) matrix and permute along axis 1 with rng.permuted(..., out=...). The statistic is a difference of means, so the treated sum alone determines it and the whole null is one matrix-vector product. Use the two-sided p-value (1 + count(|stat_perm| >= |stat_obs|)) / (B + 1); the plus-one on each side is not cosmetic, it keeps the p-value away from exactly zero and keeps the test valid at finite B.
For CUPED, fit theta = cov(y, x) / var(x) on the pooled data and use that same theta for the observed statistic and every relabelling. Pooled theta, like the pooled cap, carries no label information, so where in the loop you compute it is again only a performance question; fitting theta within arms is what goes wrong, because the adjusted outcome then depends on the labels and an observed-label fit frozen across all 20,000 relabellings breaks the match between observed and permuted statistics. x must be measured entirely before launch, which consumption_pre is. Expected variance reduction is about 1 - corr(y, x)^2; measure the achieved reduction from the two null distributions rather than asserting it.

Worked solution (45 min)

cap_y = np.quantile(df.consumption_28d, 0.99); y = np.minimum(df.consumption_28d.to_numpy(float), cap_y); cap_x = np.quantile(df.consumption_pre, 0.99); x = np.minimum(df.consumption_pre.to_numpy(float), cap_x)
t = (df.arm == 'treatment').to_numpy(); n1 = int(t.sum()); n0 = len(t) - n1; obs = y[t].mean() - y[~t].mean()
rng = np.random.default_rng(11); L = np.tile(t.astype(np.int8), (20_000, 1)); rng.permuted(L, axis=1, out=L); s1 = L @ y; stats = s1/n1 - (y.sum() - s1)/n0
p = (1 + int(np.sum(np.abs(stats) >= abs(obs)))) / (20_000 + 1)
theta = np.cov(y, x, ddof=1)[0,1] / np.var(x, ddof=1); y_adj = y - theta*(x - x.mean()); repeat steps 2 to 4 on y_adj and compare stats.std(ddof=1) between the two runs.

Expected result: A two-sided p-value strictly between 1/20001 and 1, an observed effect expressed in capped billable units per account, and a CUPED null whose standard deviation is smaller than the unadjusted one by roughly sqrt(1 - corr(y, x)^2). The CUPED point estimate stays close to the unadjusted one, since the adjustment removes variance rather than shifting the effect.

Verify it yourself

The permutation null is centred near zero: |mean(stats)| is a small fraction of std(stats).
An A/A run, relabelling arm at random before the test and repeating over seeds, gives roughly uniform p-values with about 5% below 0.05.
std(stats_cuped) < std(stats) whenever corr(y, x) is materially non-zero, and the ratio matches sqrt(1 - r^2) to within simulation noise.
Moving the pooled cap computation inside the permutation loop reproduces the same p-value bit for bit at the same seed — the direct demonstration that a pooled cap carries no label information — while setting the cap to infinity reproduces the uncapped test exactly, so the winsorisation switch is independently verifiable.

Follow-up

The p-value is 0.04 with the cap and 0.31 without it. What do you report, and what did you pre-register?
Colleagues in a shared workspace can see the treated behaviour. How does that change the design and the estimate?
How many accounts would you need to detect a 5% lift given this outcome's distribution?

Write a function to compute the distance between two geographical coordinates.

Medium · sql and data manipulation

Approach

Handle the rows that do not match: a LEFT JOIN with a NULL check is usually the question.
State the window function and its partition and ordering out loud before writing it.
Compute rates by summing numerator and denominator separately, never by averaging rates.

Follow-up

How would you verify this result without re-running the same query?
How does the query change if the join becomes one-to-many?

Can you explain the time complexity of your solution for a given problem?

Medium · sql and data manipulation

Approach

Say which table is the grain you start from, and join outward from it.
Compute rates by summing numerator and denominator separately, never by averaging rates.
Check whether any join is one-to-many before aggregating, or the sums inflate.

Follow-up

How would you verify this result without re-running the same query?
What breaks if events arrive late or out of order?

Running commitment burn-down and the date consumption crosses it

Medium · window functions · running totals · committed consumption

fct_subscription_period gives committed_amount_cents, term_start_date and term_end_date for each account's current version where pricing_model = 'committed_consumption'. fct_usage_daily gives account_id, workspace_id, sku_code, usage_date and net_amount_cents, at one row per workspace and SKU per day. Inside each account's current term, return the running total of net_amount_cents by usage_date, the first usage_date on which that running total reaches committed_amount_cents, and the fraction of the term elapsed at that point. Accounts that have not reached their commitment must still appear, with a null crossing date.

Approach

Collapse usage to one row per (account_id, usage_date) first. The fact is grained by workspace and SKU, so a raw running total leaves several rows per date and the first-crossing date becomes dependent on the arbitrary order of rows inside that day.
Restrict to the term with usage_date BETWEEN term_start_date AND term_end_date on the account's current row, so no prior term's consumption leaks into this term's burn-down.
Compute SUM(net_cents) OVER (PARTITION BY account_id ORDER BY usage_date ROWS BETWEEN UNBOUNDED PRECEDING AND CURRENT ROW). Name ROWS explicitly: the default frame is RANGE, which includes all peer rows at the same ORDER BY value and would hand back the whole day's total on each of that day's rows.
Extract the crossing with MIN(usage_date) FILTER (WHERE running_cents >= committed_amount_cents) grouped per account, which returns NULL for accounts still under commitment instead of dropping them.
Elapsed fraction is (crossing_date - term_start_date)::numeric / NULLIF(term_end_date - term_start_date, 0); Postgres date subtraction yields whole days, so the guard matters for same-day terms.
Exclude the trailing days still inside the metering settling window, measured from first_written_at against restated_at, and say how many days you cut and why.

Worked solution (30 min)

Build terms: SELECT account_id, committed_amount_cents, term_start_date, term_end_date FROM fct_subscription_period WHERE is_current AND pricing_model = 'committed_consumption' AND committed_amount_cents IS NOT NULL.
Build daily: join fct_usage_daily to terms on account_id with usage_date inside the term, then GROUP BY account_id, usage_date summing net_amount_cents.
Add running_cents with SUM(...) OVER (PARTITION BY account_id ORDER BY usage_date ROWS BETWEEN UNBOUNDED PRECEDING AND CURRENT ROW).
Aggregate per account with MIN(usage_date) FILTER (WHERE running_cents >= committed_amount_cents) AS crossing_date, then compute the elapsed fraction with the NULLIF guard.
LEFT JOIN that summary back onto terms so every committed account appears, and verify the last running_cents per account equals a plain SUM over the term.

Expected result: A per-account-per-day series of running_cents inside the current term, plus one summary row per committed account carrying crossing_date and term_elapsed_fraction, with crossing_date NULL exactly for accounts whose final running total is below committed_amount_cents.

Verify it yourself

The maximum running_cents per account equals SUM(net_amount_cents) over that account's term computed independently.
crossing_date is NULL if and only if the final running total is below the commitment.
term_elapsed_fraction falls in [0, 1] for every non-null crossing date.
running_cents is non-decreasing within each account, which fails immediately if a negative credit row was included without comment.

Follow-up

Turn this into an end-of-term overage forecast. What breaks if you extrapolate a linear run rate on a consumption product?
An account crosses its commitment at 40 percent of the term. Is that an expansion signal or a billing-surprise risk, and what would you check to tell them apart?

What metrics would you use to measure the success of an autonomous vehicle's navigation system?

Medium · product sense and metrics

Approach

Fix the population and the time window before naming any metric.
Restate the decision this analysis has to support, and who acts on the answer.
State what result would change your recommendation, so the answer is falsifiable.

Follow-up

What would you do if the primary metric and the guardrail moved in opposite directions?
Which segment would you cut first, and what would that rule out?

How do you prioritize tasks when working on multiple projects?

Medium · product sense and metrics

Approach

Decompose the metric into the rates that drive it, and say which one you would check first.
State what result would change your recommendation, so the answer is falsifiable.
Restate the decision this analysis has to support, and who acts on the answer.

Follow-up

How would you detect that the metric is being gamed rather than genuinely improving?
Which segment would you cut first, and what would that rule out?

How would you approach designing a system to track the movements of autonomous vehicles in real-time?

Medium · product sense and metrics

Approach

Name one primary metric, then the guardrail that stops it being gamed.
Fix the population and the time window before naming any metric.
Restate the decision this analysis has to support, and who acts on the answer.

Follow-up

Which segment would you cut first, and what would that rule out?
What would you do if the primary metric and the guardrail moved in opposite directions?

Given a dataset of vehicle telemetry, how would you identify anomalies in the data?

Medium · product sense and metrics

Approach

Restate the decision this analysis has to support, and who acts on the answer.
Fix the population and the time window before naming any metric.
Decompose the metric into the rates that drive it, and say which one you would check first.

Follow-up

How would you detect that the metric is being gamed rather than genuinely improving?
What would you do if the primary metric and the guardrail moved in opposite directions?

Can you explain the difference between supervised and unsupervised learning?

Medium

Approach

Say what you would check first and why it is the highest-information step.
State your assumptions explicitly before working the problem.
Clarify what is being asked and what a complete answer would contain.

Follow-up

How would you know your answer was wrong?
What assumption would you test first?

Design a seat-utilisation metric before enforcing licensed seat limits

Medium · seat utilisation · denominator design · service accounts · renewal timing

Seat limits are currently unenforced. Leadership wants a metric to decide whether to enforce them at renewal. You have dim_user_membership (user_id, account_id, role, seat_type, is_service_account, activated_at, deactivated_at, last_seen_at) and fct_subscription_period (account_id, contracted_seats, pricing_model, term_end_date, arr_cents, is_current). Define licensed seat utilisation at the account grain, state every exclusion in both the numerator and the denominator, and say what a value above 1.0 means. Deliver the metric plus the guardrail that protects against enforcement destroying more ARR than it recovers.

Approach

Numerator: distinct user_id in dim_user_membership with seat_type = 'licensed_paid', is_service_account = false, deactivated_at null and last_seen_at inside a trailing 28 days. Service accounts hold a membership and burn API keys but consume no human attention, so counting them inflates utilisation hardest on the most deeply integrated accounts, which is exactly backwards.
Denominator: contracted_seats from the fct_subscription_period version current at the evaluation date. Exclude accounts with pricing_model = 'pure_consumption' entirely, because contracted_seats is null there and the ratio is undefined rather than zero. A null coerced to zero produces an infinite ratio that will lead the report.
Interpret both tails and say what each means commercially. Above 1.0 means the entitlement is not enforced and humans are sharing seats, which is recoverable revenue. Persistently below roughly 0.5 is the most reliable available signal of a seat reduction at renewal, because buyers audit seat counts at renewal and at no other moment.
Guardrail: ARR at risk from enforcement, the sum of arr_cents over accounts with utilisation above 1.0 whose term_end_date falls inside the enforcement window. Enforcing on an account that responds by cutting seats at renewal converts a recoverable overage into a permanent downgrade, and the recovered revenue can be smaller than the loss.
Report the metric as a distribution with the two tails counted separately and the renewal calendar attached, because no account can act on enforcement until its own term ends. A fleet average of seat utilisation is a number nobody can act on.

Worked solution (20 min)

Build the current contract per account from fct_subscription_period where is_current = true, keeping contracted_seats, pricing_model, term_end_date and arr_cents, and drop pure_consumption rows.
Count active humans per account from dim_user_membership under the four filters, using distinct user_id per account_id, which is safe because the table is one row per membership and not one row per human.
Compute the ratio and bucket accounts into below 0.5, 0.5 to 1.0, and above 1.0, joining arr_cents and term_end_date.
Report ARR in each bucket and the share of that ARR renewing in the next two quarters.

Expected result: Three utilisation buckets with ARR and renewal timing attached, and a metric definition whose numerator excludes service accounts, guests, free collaborators and deactivated memberships, and whose denominator is contracted_seats from the contract version current at the evaluation date, with pure-consumption accounts excluded rather than counted as zero.

Verify it yourself

No account with pricing_model = 'pure_consumption' may appear, with a null or an infinite ratio.
Bucket ARR must sum to the total ARR over included accounts.
Summing per-account active-user counts across the fleet must be greater than or equal to the fleet-wide count of distinct user_id, because one person can hold memberships in several accounts; exact equality means multi-account users were deduplicated at the wrong grain.

Follow-up

last_seen_at is null for service accounts by definition. If the is_service_account flag is unreliable, what happens to your numerator, and how would you audit the flag?
Why a 28-day activity window rather than 7 or 90, and what evidence would change your answer?

Error rate halves while severe support tickets double

Hard · missing data · aggregation bias · reconciliation

The fleet-wide customer-visible error rate fell from 1.8 percent to 0.9 percent, while sev1 and sev2 tickets and reopened_count rose across the same fortnight. Using fct_api_request (account_id, environment, sdk_name, traffic_class, http_status, request_at, billable_units), fct_usage_daily (account_id, sku_code, usage_date, billable_quantity) and fct_support_ticket (account_id, severity, opened_at, reopened_count, linked_incident_id), determine whether reliability improved, and if not, identify precisely which rows are missing and from when. Deliverable: a diagnosis backed by an independent corroborating source.

Approach

Distrust an improvement that contradicts an independent operational signal. Two sources disagreeing is itself the finding; decide which one is more likely to be broken before explaining either.
Recompute the rate as the metric tree defines it, per account first and then as the share of accounts above the reliability target. A single global average is dominated by whichever account sends the most traffic, so a fleet number can fall while a quarter of accounts get worse.
Audit for missingness rather than for badness: count fct_api_request rows per hour split by environment, sdk_name and status class, indexed against the trailing same-hour baseline. A partial ingestion failure shows as a step drop confined to one slice, not a uniform decline.
Reconcile against a source the request pipeline does not feed, such as implied request volume from fct_usage_daily for the same accounts and dates. If the usage table is flat while request rows fell, rows are missing rather than traffic.
Test whether the missingness is differential by status, which is the mechanism that fakes an improvement: if 5xx rows are written on a path that stopped while 2xx rows were unaffected, the numerator falls faster than the denominator and the rate drops with nothing improving.
Close with the affected window, the affected slice and a restated series marked unreliable across that window, rather than a silently patched number.

Follow-up

The missing rows are unrecoverable. How do you present that fortnight in a series people compare week over week?
What monitor would have caught this within an hour, and what is its false-positive cost on a normal quiet weekend?

Suggested practice checklist

For someone who has spent the last year in notebooks, dashboards or modelling work and has not written raw SQL under time pressure. The first four days rebuild query fluency against a fixture you control and can verify by hand; the last three attach that fluency to the rest of the loop.

Day 1 — Build a fixture you can check answers against

Create a local Postgres or SQLite database with four tables (users, sessions, events, orders) holding roughly 200 rows you generated yourself, so you know the contents well enough to predict every result.
Deliberately seed the cases that break queries: a user with no sessions, a session with no events, two orders sharing a timestamp, a NULL in one join key, and one duplicated user row.
Before writing any SQL, hand-compute five answers on paper (how many users placed at least one order, median orders per ordering user, and three others) and save them as the ground truth for the week.

Deliverable: A one-command seed script plus a text file of five hand-computed answers to grade every later query against.

Day 2 — Joins, filters and NULL semantics

Answer "which users have no orders" three ways (LEFT JOIN with IS NULL, NOT EXISTS, NOT IN) and confirm that the NOT IN version returns zero rows once the subquery contains a NULL, because the comparison is never TRUE.
Reproduce the LEFT JOIN that silently collapses to an inner join by putting a right-table predicate in WHERE, then fix it by moving the predicate into the ON clause, and record both row counts.
Create a fan-out bug on purpose by joining orders to order_items and summing the order total, then correct it with a pre-aggregated subquery and explain in one line which table changed the grain.

Deliverable: One annotated .sql file holding the three join traps, each with the wrong result and the corrected result side by side.

Day 3 — Window functions and frames

Write three window queries against the fixture: a running order total per user, the rank of each order within its user by value, and the day gap to that user's previous order, then check each against the day-one ground truth.
Run ROW_NUMBER, RANK and DENSE_RANK over a column containing ties, print all three side by side, and write one sentence on when each is the correct choice.
Switch one query from the default frame (RANGE BETWEEN UNBOUNDED PRECEDING AND CURRENT ROW, which is what you get when ORDER BY is present and no frame is written) to ROWS BETWEEN UNBOUNDED PRECEDING AND CURRENT ROW, and explain why the output differs only when the ORDER BY column has duplicates.

Deliverable: Three verified window queries plus a short note explaining the RANGE versus ROWS difference in your own words.

Day 4 — The four analytical query patterns

Write a monthly retention grid: first order month per user, then months-since-first as the column, and verify that month zero equals the cohort size exactly.
Sessionize the events table under a 30-minute inactivity rule using LAG plus a cumulative sum over a new-session flag.
Build a four-step funnel that counts distinct users rather than events at each step, and state the rule you applied to a user who reaches step three without ever logging step two.

Deliverable: One file with the retention, sessionization and funnel patterns, each carrying a one-line note on the assumption it bakes in.

Day 5 — Write SQL the way you will have to write it live

Set a 12-minute timer and solve three medium prompts in a plain editor with no execution and no autocomplete, then run them and tally syntax errors separately from logic errors.
Narrate one solution aloud while writing it, stating the grain of each intermediate result (one row per user, one row per user-day) before you type its body.
Rewrite your slowest solution as a CTE chain where every CTE name states its grain, and time yourself re-solving it from blank.

Deliverable: A recording of one narrated solution plus an error tally that separates syntax from logic.

Day 6 — One day for everything that is not SQL

Write the preconditions of the two-sample t-test from memory, then check them: independent observations, and a difference in means whose sampling distribution is approximately normal, which at large sample sizes follows from the central limit theorem rather than from normality of the raw values.
Write the difference between an odds ratio from logistic regression and a relative risk, and state the condition under which the two are close (low outcome prevalence).
Prepare a 90-second answer to "how would you know this model is any good" that names the metric, the baseline you would beat, and the cost of the errors you care about.

Deliverable: One page of notes covering test preconditions, the odds-ratio caveat and the model-quality answer.

Day 7 — Full loop rehearsal

Run a 45-minute mock with someone willing to interrupt: 20 minutes of SQL, 15 minutes defining a metric, 10 minutes on a past project.
Re-solve from blank the two queries you were slowest on this week and compare the times against day five.
Write a five-line answer to "walk me through a project" that puts a number in the first sentence and names the decision the work changed.

Deliverable: Mock feedback notes plus a timed project narrative you can deliver without reading it.

Being wrong in public

A number you shipped turned out to be wrong, and someone had already acted on it. That is one of the most useful stories a data person can carry. What is being scored is how fast you noticed, who you told first, and what you changed in the process so the same class of error could not repeat quietly.

Can you provide an example of how you handled a conflict within a team?

Medium · behavioural and stakeholder questions

Approach

Quantify the outcome, including what you would not claim credit for.
Close with what you would do differently, concretely.
Pick a story where you drove the decision, not one where you observed it.

Follow-up

What would you do differently if you ran that project again?
What did you decide not to do, and why?

Disagree with a product manager about an adoption claim

Medium · conflict · telemetry hygiene · adoption metrics

A product manager is about to present that a new SDK release drove a 40 percent rise in requests among adopting accounts, computed from fct_api_request counts grouped by sdk_version. You find the rise is concentrated in traffic_class equal to ci, that rows with is_retry true grew alongside it, and that restricting to interactive non-retry traffic leaves a 3 percent lift. The launch review is in two days. Decide how you raise this, with whom and in what order, and what you propose the claim becomes.

Approach

The interviewer is probing whether you can correct a colleague without ambushing them, and whether your own counter-analysis carries the caveats theirs lacked. Go to the product manager privately before the review. A correction delivered in the room is a status move and loses the argument you are actually trying to win.
Bring a decomposition rather than a verdict: the same accounts and window, requests split by traffic_class with retries held out as their own column, so their 40 percent and your 3 percent reconcile line by line and neither has to be taken on trust.
Reproduce their figure exactly first. If you cannot land on 40 percent with their method, you do not yet know what you are disagreeing with.
Ask whether the continuous-integration lift is itself valuable. An account wiring the SDK into its pipeline has increased integration depth, which is the dominant switching cost in this domain, so the honest claim may be that integration depth rose while interactive usage moved 3 percent. Improving the claim beats deleting it.
Name the mechanism that makes the raw count dangerous: clients retry when the platform degrades, so retry volume climbs exactly when the customer is most at risk. Pull the 5xx rate for the same accounts and window before anyone concludes anything, and note that billable_units is zero on 5xx rows, so request counts and billable quantities diverging is itself the signal.
Close with a standing definition for launch metrics so the next release does not repeat the exercise.

Follow-up

The product manager argues that continuous-integration traffic is real usage and declines to split it out. Is that position defensible, and under what metric definition?
Suppose the 5xx rate for those same accounts also rose 40 percent. What is the claim now?
The review happens and the raw number is presented regardless. What do you do next, and what do you not do?

Scope an open-ended request to predict account churn

Medium · scoping · churn modelling · operating point

A customer success director asks for a list of accounts about to churn. You know only that the team has six people and that contracts are annual. Available data is fct_subscription_period, fct_usage_daily, fct_api_request, fct_support_ticket and dim_account. Before writing any code, produce the questions you need answered, a proposed definition of about to churn, and the shape of the artefact you would hand back, including the operating point that turns a score into a decision.

Approach

The interviewer is probing whether you convert a vague request into a decision with a capacity constraint attached. A candidate who starts talking about model families has already failed the exercise.
Pin the event and the horizon first. Churn is only possible at term_end_date, so the population is accounts renewing in the next 60 to 90 days, not the whole base. Ask explicitly whether contraction and downgrade count as churn or only full non-renewal, because the three have different base rates and different interventions.
Pin the action and the capacity. Six people times a realistic number of meaningful interventions per week gives k, and k is what the list is ranked to. Evaluate on precision at k rather than a global AUC over accounts that will never be contacted.
Audit leakage before choosing features. Every feature needs a timestamp proving it existed before the prediction date. A downgrade amendment, a churn reason code, and a ticket opened after the renewal conversation started are all leaks that will make the offline number look excellent and the live list useless.
Ask for the counterfactual now rather than later. Coverage is assigned deliberately, so without a held-out slice agreed at the start the intervention can never be evaluated, and you will be asked for its impact in nine months regardless.
Propose the smallest artefact that closes the loop: a weekly ranked list sized to capacity with two or three inspectable reasons per row, plus a stated policy for accounts below the line.

Follow-up

The director insists all accounts are in scope, not only those renewing soon. How do you answer without simply refusing?
Historical non-renewals number about 30 a year. At what point do you tell them a model is the wrong tool and a rules list is better?
Which candidate features would you drop purely because you cannot date them?