At Retina AI, a Data Scientist is at the absolute core of the company's product offering and business value. Retina AI specializes in early-stage Customer Lifetime Value (CLV) predictions, helping brands optimize their customer acquisition, retention strategies, and marketing budgets. As a Data Scientist here, you are not just building models in a vacuum; you are directly developing the proprietary machine learning algorithms that drive revenue and strategic decisions for enterprise clients.
This role requires a unique blend of deep statistical knowledge, predictive modeling expertise, and strong business acumen. Because Retina AI's core product is data science itself, your work has a direct impact on the company's reputation and product capabilities. You will work on complex, high-dimensional datasets to predict customer behavior, requiring you to balance mathematical rigor with scalable engineering practices.
You will collaborate closely with engineering, product, and client success teams to integrate predictive models into production systems and translate complex data outputs into actionable business strategies. For anyone passionate about customer analytics, Bayesian statistics, and seeing their models directly influence business bottom lines, this position offers an incredibly high-impact and intellectually stimulating environment.
Recruiter Phone Screen
reportedData 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 Discussion
reportedMuch of what gets scored here happens out loud while you type. Nobody can see your reasoning inside a half-written query, so five silent minutes read as being stuck even when they are not. State the plan in plain language first: which tables, what grain you are aggregating to, and the one filter that defines the population. Then write it. The narration doubles as insurance, because a wrong plan gets caught early and cheaply while a wrong query gets caught at the end with no time left to redo it. A timed statistics section, where one exists, is a separate test with its own clock.
What to demonstrate
- Whether the query you write matches the plan you just described
- What you do with a hint, meaning whether the correction gets absorbed or the first approach gets defended
- Whether you can debug your own wrong output by reading the result set and naming which part of the query produced the anomaly
How to prepare
- Solve three problems while screen-sharing into a recording, then watch it back and mark every stretch longer than thirty seconds where you said nothing
- Practise compressing the plan into one sentence before typing, then check afterwards whether the finished query actually matched it
- Time yourself on statistics questions that carry a business reading, such as what a confidence interval does and does not claim, rather than re-reading notes without a clock
Take-Home Challenge
reportedYour submission is read asynchronously by someone who cannot ask you a clarifying question, and who will usually skim it before reading it properly. That changes what good looks like. The conclusion belongs near the top with the supporting analysis beneath it, and the code should run start to finish on a clean machine without a manual step you forgot to document. A reviewer forced to reconstruct your reasoning from the order of notebook cells is already discounting the work. The submissions that land are the ones where a busy reader gets the answer immediately and can verify it if they want to.
What to demonstrate
- Whether the answer arrives before the methodology, so a reader who stops after the first page still has the recommendation
- Whether the code runs end to end from the submitted files, with dependencies and data paths declared rather than assumed
- Whether each chart is legible on its own, carrying axis labels and units, and exists to support a claim made in the text
How to prepare
- Restructure a past analysis so the opening paragraph holds the recommendation and the number behind it, then check that nothing later in the document quietly contradicts it
- Copy your own submission into an empty directory, run it in a clean environment, and fix everything that breaks; hidden local state is caught here or by the reviewer
- For every chart, write the one sentence it is meant to prove, and delete the chart if you cannot write that sentence
Onsite Interview
reportedA day of back-to-back interviews samples your floor, not your ceiling. Four hours in, the habits that carry a good answer are the first to go: restating the question before solving it, asking what the data would have to look like, checking a number before quoting it. What the day decides is whether the tired version of you is still someone to leave alone with an ambiguous problem. The round that sinks a candidate is usually not the hardest one. It is the one immediately after the round that went badly.
What to demonstrate
- Whether the late rounds get the same clarifying questions as the first one, or whether you start answering immediately to save effort
- Whether a weak answer stays in the room it happened in, instead of following you into the next conversation as apology or distraction
- Whether the quality of your questions holds up, since fatigue removes curiosity about the problem before it removes knowledge of the method
How to prepare
- Rehearse the length, not just the content: book four mock interviews of different types in one afternoon with short gaps, because the one you need to observe is the fourth
- Put the two or three questions you ask at the start of any problem on a card in front of you, so that under fatigue it is a habit you run rather than a decision you make
- Decide in advance what the gap between rooms is for: water, one line of notes on anything you promised to follow up, and an explicit close on the round that just ended so it does not travel
- Prepare a different closing question for each interviewer, so the end of a long day does not produce the same one four times
PracHub editorial advice for the preparation topics above.
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.
Treating raw request or usage volume as engagement
Most traffic in this domain is emitted by machines. Continuous-integration pipelines, scheduled batch jobs, synthetic monitors, backfills and client retries can all grow by an order of magnitude from one configuration change made by one engineer, and none of it represents a new decision to use the product. The inversion is what makes it dangerous: when the platform degrades, clients retry, so error-driven retry volume rises at the exact moment the customer is most likely to leave, and an engagement dashboard built on raw counts shows growth immediately before a churn. Filter on traffic_class and on successful status before anything else, and keep failed-request volume as its own separate series.
Naming a model class before naming the deployment constraints
Set out the latency budget, the label delay, the retraining cadence, the interpretability requirement and the number of labelled examples, then pick the model that fits them. A boosted-tree answer to a problem where each decision must be explained to the affected user is a well-executed answer to the wrong question.
SQL that silently fans out on a one-to-many join
State the grain of each table and the grain you want in the result before writing the join. Pre-aggregate the many side to the join key, or use EXISTS or a window function, and verify with a row count against COUNT(DISTINCT id) rather than trusting that the numbers look plausible.
Choose a category, try a prompt, then open its approach, worked solution or follow-up when you need it.
Explain the difference between Frequentist and Bayesian approaches to …
Explain the difference between Frequentist and Bayesian approaches to probability, and when you would choose one over the other.
Approach
- Say what the estimate is of, and over what population it generalises.
- Translate the result into the decision it informs, in one plain sentence.
- Sanity-check the answer against a simple bound or a simulated case.
Follow-up
- Which assumption here is most likely to be violated in practice?
- What sample size would you need to detect an effect half this size?
How would you measure the performance of a Customer Lifetime Value (CL…
How would you measure the performance of a Customer Lifetime Value (CLV) model when historical data is limited?
Approach
- Pick an evaluation metric that matches the cost of each error type, not a default.
- Check what information would not exist at prediction time, and exclude it.
- Say how the offline result would be validated online before it is trusted.
Follow-up
- What would you monitor after launch to know the model is still valid?
- Where could label leakage enter this setup?
Describe a time you had to optimize a model's hyperparameters. What wa…
Describe a time you had to optimize a model's hyperparameters. What was your approach, and how did you prevent overfitting?
Approach
- Frame the prediction: the label, the moment of prediction, and the action it triggers.
- Say how the offline result would be validated online before it is trusted.
- Pick an evaluation metric that matches the cost of each error type, not a default.
Follow-up
- How would you choose the decision threshold, and who owns that choice?
- Where could label leakage enter this setup?
Explain the trade-offs between using a traditional regression model ve…
Explain the trade-offs between using a traditional regression model versus a survival analysis approach for customer churn.
Approach
- Pick an evaluation metric that matches the cost of each error type, not a default.
- Set a baseline first, so any model has something honest to beat.
- Check what information would not exist at prediction time, and exclude it.
Follow-up
- Where could label leakage enter this setup?
- How would you choose the decision threshold, and who owns that choice?
Quantify billable volume created by retries after server errors
From fct_api_request (request_id, account_id, endpoint, idempotency_key, is_retry, http_status, request_at, billable_units), measure how much billable volume in a 28-day window is retry traffic that followed a 5xx. Group requests into attempt chains by (account_id, endpoint, idempotency_key) ordered by request_at; a request is error-driven if any earlier attempt in its chain returned 5xx. Requests with a null idempotency_key cannot be chained, so report them as their own class rather than assuming each is unique. Return billable_units split into first-attempt, error-driven retry, other retry and unchainable, per account.
Approach
- Split the population before measuring anything. A null idempotency_key is not a chain of one, it is an unknown; report its share of billable_units first, because if it is 40% of volume then the headline estimate is a lower bound and the deliverable has to say so.
- Within chainable rows, sort by (account_id, endpoint, idempotency_key, request_at) and derive 'any earlier attempt failed' with arithmetic rather than a per-group lambda: with is5 = (http_status >= 500), the per-chain cumsum minus the row's own value is positive exactly when an earlier attempt in that chain returned 5xx. A groupby-apply gives the same answer and is unusable at five million rows.
- Do not take is_retry as the definition. It is set by the client whenever an idempotency_key is resent, which covers retries after client-side timeouts and after 4xx as well; compute the flag yourself and then cross-tabulate it against is_retry, because the disagreement is a finding in its own right.
- Aggregate billable_units by (account_id, class) and assert the classes sum to each account's total. The spine sets billable_units to zero on 5xx responses, so the failed attempt contributes nothing and the whole inflation sits in the successful retry that follows it.
- Report the per-account share and look at its distribution, not the fleet total. One account in a retry storm dominates any blended figure, which is the same failure that makes a fleet-wide error rate useless.
Worked solution 40 min
- unchainable = df.idempotency_key.isna(); report df.loc[unchainable].groupby('account_id').billable_units.sum() before proceeding.
- keys = ['account_id','endpoint','idempotency_key']; c = df[~unchainable].sort_values(keys + ['request_at'], kind='mergesort'); c['is5'] = (c.http_status >= 500).astype(int)
- c['attempt_no'] = c.groupby(keys, sort=False).cumcount(); c['prior_5xx'] = (c.groupby(keys, sort=False).is5.cumsum() - c.is5) > 0
- c['cls'] = np.where(c.attempt_no == 0, 'first_attempt', np.where(c.prior_5xx, 'error_driven_retry', 'other_retry')); out = pd.concat([c, df[unchainable].assign(cls='unchainable')]).groupby(['account_id','cls']).billable_units.sum().unstack(fill_value=0)
Follow-up
- An account's error-driven share is 22%. Is that the platform's fault or the client's, and what do you look at next?
- How would you define a consumption-based north-star metric that an outage cannot inflate?
- Chains straddle the 28-day boundary. How large is that bias and in which direction?
Running commitment burn-down and the date consumption crosses it
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.
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?
Find paid seats that never called the API
dim_user_membership carries user_id, account_id, seat_type, is_service_account, deactivated_at and last_seen_at. fct_api_request carries user_id, which is null whenever the caller is a service account or an unattended key, plus account_id, request_at and http_status. For a single account, list every licensed_paid seat with is_service_account = false and deactivated_at null that issued no request at all in the trailing 90 days, returning user_id and last_seen_at. A first draft writes NOT IN against a subquery over fct_api_request.user_id. State exactly what that draft returns and why, then write the correct query.
Approach
- Read the column comment first: user_id is nullable on the fact, so the subquery almost certainly contains at least one NULL for any account that runs a service account or an unattended key.
- Work the three-valued logic out loud. x NOT IN (a, NULL) expands to NOT (x = a OR x = NULL); the second disjunct is UNKNOWN, so for any x not equal to a the whole predicate is UNKNOWN and the row is filtered out. The draft returns zero rows, which reads as full seat utilisation.
- Rewrite as NOT EXISTS with the 90-day and status predicates inside the correlated subquery. Putting them in the outer WHERE instead turns the anti-join into a different question and silently changes the answer.
- Correlate on both user_id and account_id. A membership is (user_id, account_id) and one person can hold memberships in several accounts, so correlating on user_id alone marks a seat as active because that human was busy somewhere else.
- Filter the seat side to seat_type = 'licensed_paid', is_service_account = false and deactivated_at IS NULL, and say what the resulting count means next to contracted_seats on the current subscription row.
Follow-up
- Adding AND user_id IS NOT NULL to the subquery also fixes NOT IN. Why is NOT EXISTS still the form you would leave in the repository?
- last_seen_at looks like a shortcut for the whole query. What does it actually record, and where does it disagree with API activity?
- This is a seat-reduction risk list. What threshold would you attach before handing it to an account team, and what happens to seats below it?
How do you prioritize your work when faced with multiple urgent reques…
How do you prioritize your work when faced with multiple urgent requests from different teams?
Approach
- Decompose the metric into the rates that drive it, and say which one you would check first.
- Name one primary metric, then the guardrail that stops it being gamed.
- Fix the population and the time window before naming any metric.
Follow-up
- What would you do if the primary metric and the guardrail moved in opposite directions?
- How would you detect that the metric is being gamed rather than genuinely improving?
How do you design an A/B test to measure the lift of a new marketing c…
How do you design an A/B test to measure the lift of a new marketing campaign when customer purchase cycles are long?
Approach
- Name the randomisation unit first; it decides the variance and what the test can detect.
- Name the guardrails that would stop a launch even on a positive primary result.
- Say whether units interfere with each other, and switch design if they do.
Follow-up
- What would you conclude if the result is positive but the test is underpowered?
- What would you do if you could not randomise at all?
Build an activation metric tree for a self-serve onboarding rebuild
The self-serve signup flow is being rebuilt. You have dim_account (account_id, created_at, is_internal, first_workspace_at, is_current) and fct_api_request (account_id, workspace_id, environment, api_key_id, http_status, traffic_class, request_at). Build a three-layer metric tree from signup through to production dependency, name the primary metric for this specific change, and state the fixed measurement window and the denominator for each layer. Two of the three layers share a denominator and the third does not, so say which invariant still holds across the tree once the denominators differ. Say which layer the onboarding team can actually move and which it cannot, and why raw signup counts are not on the tree.
Approach
- Fix the cohort once: accounts created in cohort period P with is_internal = false, collapsed to one row per account_id since dim_account is type 2. Say out loud that employee, demo and load-test accounts are removed, otherwise the tree partly measures internal testing.
- Layer one, qualified signup, on the cohort denominator: accounts recording at least one successful request (http_status < 400) with traffic_class in ('interactive','batch') within 14 days of created_at. This replaces raw signup counts, which measure registration-form friction and duplicate sign-ups from one organisation rather than demand.
- Layer two, activation, also on the cohort denominator: the share of the cohort whose first successful request with a non-null api_key_id arrives within a fixed 168 hours of created_at. Fixed window, not trailing, so that cohort weeks are directly comparable and no cohort is censored.
- Layer three, production depth, on a conditional denominator: the share of activated accounts holding a workspace with environment = 'production' whose trailing 7-day count of successful non-synthetic requests exceeds a floor calibrated from the data, for example the volume above which 90-day retention stops rising. A floor picked by intuition makes this a metric about the floor. Conditioning on activation is the right operational choice, because the question it answers is whether accounts that got started go on to depend on the product, but it means this rate is not on the same scale as the two above it: a conditional rate of 0.50 sits perfectly happily underneath a 0.30 activation rate. To place it back on the funnel, multiply it by the activation rate and report that product as the cohort share.
- Align the traffic_class filter across all three layers, otherwise the underlying account sets do not nest: an activation definition that admits traffic_class = 'ci' will count accounts that layer one excluded, and the tree stops being a funnel. With mixed denominators the invariant that survives is on account sets and therefore on counts, not on rates. Then name the primary metric for this change, the seven-day activation rate, because the onboarding flow controls the path to first successful call and nothing below it. Production depth is the lagging check that the activation gain was real rather than one sample request.
Worked solution 25 min
- Build the cohort table: one row per account_id, bucketed by the ISO week of created_at, is_internal = false, dim_account collapsed to is_current = true.
- For each layer, find the first qualifying request per account from fct_api_request under that layer's filter and keep min(request_at), using EXISTS or a lateral so the cohort stays at one row per account.
- Compute layers one and two as shares of the cohort denominator and layer three as a share of the activated accounts, carrying the raw account count beside every rate, and mark any cohort week not yet clear of the longest window (14 days) as not reportable.
- Plot the three rates by cohort week, and separately confirm nesting on the counts within each cohort.
Follow-up
- Activation rises 6 points and the production-depth rate among activated accounts is flat 60 days later. Say what happened to the cohort share of production-depth accounts, and why a flat conditional rate is the better of the two readings available here.
- Median time-to-first-successful-call answers the same question with more information. Why is it a worse weekly dashboard metric, and what estimator would make it honest?
Billable units per account jumped while nothing shipped
Billable units per paying account rose 22 percent month over month with no pricing or packaging change. From fct_api_request (account_id, endpoint, http_status, is_retry, idempotency_key, traffic_class, billable_units, request_at) and fct_usage_daily (account_id, sku_code, usage_date, billable_quantity), determine how much of the rise is delivered value and how much is duplicated work. Deliverable: the decomposed figure, the accounts it concentrates in, and a recommendation on whether to report the 22 percent at all.
Approach
- Check the guardrail before the headline: compute the customer-visible server error rate per account for both months, numerator http_status >= 500 and denominator excluding synthetic_monitor and load_test. Metered volume rising alongside an error rate is the known failure mode in this domain.
- Deduplicate logical work by counting billable_units once per (account_id, idempotency_key) at the first successful request rather than once per row. Rows with a null idempotency_key cannot be deduplicated, so report their share as an explicit uncertainty band instead of assuming they are all unique.
- Split by traffic_class before interpreting anything, because one change to a continuous-integration configuration can multiply request volume overnight without a human deciding anything about the product.
- Test concentration: compute the per-account distribution of the increase and its top-decile share. A rise carried by a few accounts scaling one batch job is a different finding from a broad shift and gets a different recommendation.
- Reconcile against fct_usage_daily for the same accounts and dates, and be ready to explain the expected gap in two sentences: request rows include retries and failures carrying zero billable_units, and the usage table restates after first write.
Follow-up
- A client retrying a request the server already completed produces duplicate billed work. What protocol or product change removes that, and what would you measure to confirm it worked?
- If the duplicated volume was genuinely invoiced, what should finance do, and how does that change what belongs in the metric?
For a candidate whose interviews will centre on A/B testing, metric movement and causal claims. Design comes before arithmetic, arithmetic before analysis, and the week ends by rehearsing the readout rather than the derivation.
Prepare, practise & reflect
One practical outcome each day. Spend longer where you need it.
0 / 7 done01Design one test end to end on paper
- Take a single feature change and write the full design: randomization unit, the exact point of exposure, the primary metric with its grain, guardrails, allocation, planned duration, and the decision rule committed before any data exists.
- Write why the randomization unit must sit at or above the level where treatment can spill over, and give one case where user-level randomization is still contaminated (shared accounts or devices, or two participants in the same marketplace).
- State in advance what you will do if the primary metric is flat while a secondary metric is significant.
Deliverable: A one-page test design with a decision rule written before launch.
Practice prompt ↗Practice prompt ↗Worked solution ↗02Power arithmetic until it is automatic
- Compute required sample size per arm for a binary metric with the normal approximation, n is approximately 2 times (z for alpha/2 plus z for power) squared times p(1 minus p) divided by delta squared, for baselines of 2, 10 and 40 percent at a 5 percent relative lift, and note that for a fixed relative lift the requirement falls as the baseline rises because delta grows proportionally with p.
- Redo the calculation for a continuous metric using variance in place of p(1 minus p), and show why a heavy-tailed quantity such as revenue per user needs either far more traffic or a capped version with a stated cap.
- Convert one of the results into weeks given a weekly eligible traffic figure, then list the two honest ways to shorten it (accept a larger detectable effect, or reduce variance) and write why quietly lowering the power target is a decision to miss more real wins, not a speedup.
Deliverable: A small script or sheet that maps baseline, minimum detectable effect, alpha and power to sample size and weeks, cross-checked against a published calculator.
Practice prompt ↗Practice prompt ↗03Variance and the unit-of-analysis problem
- Take a ratio metric whose denominator is not the randomization unit (clicks per session, randomized by user) and compute the standard error twice, once naively at session level and once by the delta method or a user-level bootstrap, then record how much the naive version understates it.
- Implement CUPED on simulated data: choose a pre-period covariate X measured before assignment, estimate theta as Cov(Y, X) divided by Var(X), and analyse Y minus theta times (X minus its mean) in place of Y. Confirm the variance of the adjusted outcome equals the raw variance multiplied by one minus the squared correlation between Y and X, so a correlation of 0.45 removes about 20 percent of the variance and not 80.
- Now run that simulation a few hundred times and confirm the adjusted effect estimate is unbiased for the same effect rather than numerically identical to the raw one. Within any single run the two differ, sometimes by a large fraction of the true effect, because the two arms' pre-period covariate means never coincide exactly in a finite sample; they agree in expectation, which is the property that matters and the one to state out loud.
Deliverable: A notebook showing the adjusted estimator with a measurably smaller variance than the raw one, plus a repeated-simulation table showing the two estimators agreeing on average while differing run by run.
Practice prompt ↗Practice prompt ↗04Validity threats you can actually test for
- Run a sample ratio mismatch check as a chi-square goodness-of-fit test against the intended allocation, and write the three causes you would chase first (assignment logged before exposure, an arm-specific redirect or load failure, bot filtering applied asymmetrically).
- Simulate peeking: generate A/A data, test daily at alpha 0.05 across 14 looks, record the inflated false positive rate, then apply an alpha-spending boundary or commit to a fixed horizon and confirm the rate returns to nominal.
- Write how you would separate a novelty effect from a durable lift using the treatment effect plotted against days since first exposure, and what shape would change your recommendation.
Deliverable: One table showing the peeking false positive rate before and after correction, plus a written SRM triage list.
Practice prompt ↗Practice prompt ↗Worked solution ↗05When randomization is not available
- Write the identifying assumption for difference-in-differences (parallel trends in the absence of treatment), then plot pre-period trends for two candidate control groups and justify rejecting one of them.
- Design a switchback test for a change where user-level randomization would leak across participants, choosing a time-block length against the carryover you expect and saying how you would detect carryover in the data.
- List what an interrupted time series or a synthetic control buys you and the one thing neither can rule out: an unobserved shock that coincides with the launch.
Deliverable: A one-page memo recommending a single quasi-experimental design and naming its weakest assumption explicitly.
Practice prompt ↗Practice prompt ↗06The readout query
- Write the assignment-to-exposure join that returns exactly one row per unit per experiment, and handle units appearing in both arms by excluding and counting them rather than silently keeping one.
- Compute the per-arm metric, its variance and the relative lift with a confidence interval in SQL, then reproduce the identical numbers in a notebook as a cross-check.
- Add a segment breakdown and write the sentence that keeps it from being p-hacking: segments declared in advance, everything else reported as exploratory and corrected for multiplicity.
Deliverable: A single query that outputs the full readout table, matched to a notebook recomputation.
Practice prompt ↗Practice prompt ↗07Present it to someone who will not read the appendix
- Give a 10-minute readout of a real or simulated experiment in the order decision, number, uncertainty, caveat.
- Have your listener ask "can we ship it" in the case where the primary is flat and a guardrail moved, and answer with a recommendation rather than a request for more data.
- Rewrite your opening line so the recommendation lands before any methodology.
Deliverable: A one-page readout whose first line is the recommendation.
Practice prompt ↗Practice prompt ↗Worked solution ↗Expand any day for tasks and deliverables. Your progress is saved on this device.
An answer without a quantity is hard to interrogate, so interviewers keep probing until they find one. Come with the baseline, the change, the window it was measured over, and how confident you were. If the effect never got measured, say so and say what you would have measured. Fabricated precision is worse than an honest gap.
How do you handle highly imbalanced datasets when training a predictiv…
How do you handle highly imbalanced datasets when training a predictive model?
Approach
- Name the disagreement or constraint, and how you resolved it with evidence.
- State the situation in two sentences and spend the rest on your reasoning.
- Quantify the outcome, including what you would not claim credit for.
Follow-up
- How did you know the outcome was caused by your change?
- What did you decide not to do, and why?
Share an experience where a data science project did not go as planned…
Share an experience where a data science project did not go as planned. What went wrong, and what did you learn from it?
Approach
- State the situation in two sentences and spend the rest on your reasoning.
- Name the disagreement or constraint, and how you resolved it with evidence.
- Pick a story where you drove the decision, not one where you observed it.
Follow-up
- How did you know the outcome was caused by your change?
- What did you decide not to do, and why?
Disagree with a product manager about an adoption claim
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?
- 01
How do you handle highly imbalanced datasets when training a predictive model?
- 02
Share an experience where a data science project did not go as planned. What went wrong, and what did you learn from it?
- 03
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.
Is this an official Retina AI interview guide?
No. It is PracHub's own research and practice material for the Data Scientist role at Retina AI. Rounds and questions reflect what candidates have reported, not a process Retina AI has published, and they change over time. Confirm the current format and scope with your recruiter.
PracHub interview research ↗How difficult is the Retina AI Data Scientist interview process?
Candidates generally rate the difficulty as average to difficult. The technical bars for the take-home challenge and statistical theory are high, but the interviewers are supportive and looking to see how you think through complex problems.
PracHub interview research ↗What is the company culture like at Retina AI?
The culture is collaborative, intellectually curious, and fast-paced. The team values open communication, scientific rigor, and a strong sense of ownership. The onsite team lunch is a testament to how much they value personal connection and team cohesion.
PracHub interview research ↗How quickly does the team make a decision after the final round?
The timeline can vary depending on the hiring pipeline, but candidates typically receive feedback or an update within 1 to 2 weeks following their onsite interview.
PracHub interview research ↗Is the take-home assignment graded on a specific rubric?
Yes. The team evaluates your take-home based on code cleanliness, the rigor of your validation strategy, your exploratory data analysis, and how clearly you document your assumptions and business recommendations.
PracHub interview research ↗Sources & methodology 3 sources ↗
No official company page is cited. Rounds and questions come from candidate reports and PracHub editorial material; each source shows the date it was read.
- 01PracHub interview research ↗
PracHub editorial research into this company and role, maintained with this guide. Candidate-reported, not an employer publication.
PracHub page · Accessed 2026-09-22 - 02PracHub Data Scientist practice ↗
Cross-company practice questions for this role.
PracHub page · Accessed 2026-09-22 - 03PracHub interview preparation framework ↗
The framework the preparation plan follows.
PracHub page · Accessed 2026-09-22
Explore related engineering environments and compare the preparation focus.
[24]7.ai Data Scientist Interview Questions & Guide 2026
Modeling drills for [24]7.ai data scientist prep: choosing a baseline, reading regression coefficients honestly, handling class imbalance, and defending a model you would actually ship.
Xcellent Technology Solutions Data Scientist Interview Questions & Guide 2026
Modeling drills for Xcellent Technology Solutions data scientist prep: choosing a baseline, reading regression coefficients honestly, handling class imbalance, and defending a model you would actually ship.
Wolters Kluwer Data Scientist Interview Questions & Guide 2026
Causal-inference exercises for Wolters Kluwer data scientist prep: difference-in-differences, matching, instrumental variables, and knowing when observational data simply cannot answer the question asked.
Zscaler Data Scientist Interview Questions & Guide 2026
Data-quality drills for Zscaler data scientist prep: survivorship and selection bias, leakage through time-ordered features, and extreme values where provenance decides the call. A logging fault or unit slip gets corrected or dropped; a gen