Loading...
Big Tech and product companies, senior. 31 questions across 7 blocks, about 16 hr of work.
The defining feature is calibration rather than difficulty. Every interviewer works from the same rubric and writes evidence-based feedback, so the panel is trying to place you at a level rather than decide whether you can do the job. That means the coding round is genuinely algorithmic and time-boxed, the design round widens in scope as the level rises, and a strong answer with no stated trade-offs scores worse than a middling answer that names what it gave up.
Front-load breadth, put the two hardest rounds mid-week when you still have time to react to a bad session, and keep the last evening for a timed loop and your stories. Two hours a night done every night beats one eight-hour Saturday, because retrieval improves with spacing and a single long session mostly produces the feeling of having worked.
Read the sheets for the core sections of this track at speed. Mark the items you could not answer; they become your revision list all week.
Uniform sampling needs a gapless array you can index, so a deletion cannot leave a hole and cannot shift the tail. You overwrite the hole with the last element and shrink by one, which costs a fixed amount of work but silently moves another member, so the map from value to index must be corrected for the element that moved and not only for the one removed.
Spend roughly three minutes agreeing the functional scope, three pinning the numbers that change the architecture, and four turning those numbers into a derived estimate of request rate, storage and bandwidth — then state which design decision each number forces.
One-hot is unusable at that width, so the options are frequency encoding, smoothed target encoding, hashing, or a library's native categorical handling. Target encoding is the strongest and the most dangerous, because it must be fitted inside each fold or it leaks the label. Use this categorical encoding answer to show the decision, trade-off, and evidence rather than a memorised definition.
The stack and concept questions this loop opens with. Answer before reading, and note which probes you missed rather than whether you got it right.
One queen per row is forced, so search row by row and keep sets of used columns, used sums and used differences to reject a conflict before recursing. The anti-diagonal is the row plus column and the main diagonal the row minus column, and that difference must be shifted into range, not reduced with an absolute value.
Treat errors as published contract: an HTTP status that classifies the failure, a stable machine-readable code the client branches on, a human message that carries no internal detail, per-field detail for validation, and an explicit statement of whether a retry is safe.
Build the split to imitate the gap between training and deployment. Group all of an entity's rows onto one side when rows repeat per entity, split forward in time for anything temporal with an embargo covering the label window, and treat a plain random shuffle as the default that silently inflates almost every score.
Thread-safe component design starts by assigning every field a policy: confined, immutable or guarded by a named lock. Method-level synchronization is not enough because atomic methods do not compose into atomic workflows.
A training row must contain only feature values that were knowable at its own event timestamp. A naive equality join on entity id attaches whatever the feature table holds now, including values derived from after the event, which leaks the future and produces offline metrics that collapse in production.
Whiteboard each one for real, on paper, before you look. A design answer you have only read is one you cannot produce under questioning.
Split the values across a max-heap of the lower half and a min-heap of the upper half, so the median is read off one or both tops. Balancing by size alone is what breaks it: a new value has to enter through the heap it belongs to by value and be transferred across, never pushed into whichever heap is currently smaller.
Model resources around client workflows, paginate with an opaque keyset cursor because offset pagination skips and duplicates rows under concurrent writes, make unsafe operations replayable with a client-supplied idempotency key, and evolve additively so that versioning stays a last resort rather than a routine.
Train-serve skew is a mismatch between features used in training and features produced at serving time. Prevent it with shared transformations, point-in-time joins, feature logging, schema checks and monitoring of served feature distributions. It also connects feature store to the point an interviewer is testing.
A producer-consumer bounded buffer applies backpressure to the producer instead of letting an unbounded queue absorb overload until the process dies. Deciding what a full buffer does - block, drop, or reject - is the decision that determines how the system fails.
The rounds where there is no correct answer, only a defensible one. State your assumptions out loud; that is what is being scored.
A cache stampede happens when one hot key expires and every concurrent miss rebuilds the same value. Let one request rebuild while the rest wait or serve stale, then prevent repeats with stale-while-revalidate, jittered TTLs and refresh-before-expiry. Use this caching answer to show the decision, trade-off, and evidence rather than a memorised definition.
Work backwards from the review capacity. Fix the alert budget the team can clear, evaluate precision and recall at exactly that budget on data at the real prevalence, define the label and its arrival delay before modelling, and never report accuracy, which a do-nothing model scores 99.95% on.
Pull a bounded, representative extract into an environment you control and develop against that, keeping production for read-only verification under agreed limits, and treat every write path as something proven against a restored copy before it runs anywhere near live.
The dialog was built as a styled div, so nothing marks the rest of the page inert. Focus was never moved in, never constrained, never restored on close, and Escape was never wired. A correct dialog moves focus in, keeps it in, restores it, marks background inert and is labelled - which native dialog with showModal gives you almost for free.
Escape analysis moves a value to the heap when the compiler cannot prove its lifetime ends with the frame that created it, and go build -gcflags=-m prints every decision. Which of those allocations matter is a measurement question, answered with a benchmark, an alloc_space profile and benchstat.
Return to whatever went worst this week. One session on a genuine weakness is worth three on material you already have.
The quadratic dynamic programme asks for the best chain ending at each index. The faster version keeps one array where slot j holds the smallest possible tail of an increasing subsequence of length j plus one; that array stays sorted, so each element is placed by binary search for O of n log n.
Accept the request, return 202 with the location of an operation resource, and do the work behind a queue. Clients poll that resource or receive a signed callback, and either way the operation id is the contract - one that a callback can reach before the submitter has finished recording it.
Because you cannot score a million-item catalogue per request inside a few tens of milliseconds. Retrieval reduces millions to hundreds with a cheap model judged on recall, then a far more expensive ranker orders those hundreds and is judged on the ordering it produces at the top of the list.
In a parking lot machine coding round, scope one deterministic flow of park, unpark and charge, model Vehicle, Slot, Level, Ticket and a ParkingLot facade, then push slot allocation and pricing behind strategy interfaces so review-time variants are configuration rather than surgery.
Run the rounds back to back with a timer and no notes. The point is stamina and switching cost, which no amount of single-question practice reveals.
Precompute, for each prefix of the pattern, the length of the longest proper prefix of the pattern that is also a suffix of that prefix. On a mismatch you fall back to that length instead of rewinding the text, which is what turns the quadratic naive scan into a linear one.
An idempotency key for retried charges is minted by the client per payment intent, claimed under a unique constraint in the same transaction as the charge, and used to replay the stored response on repeat. The API contract must define retention, in-flight retries and mismatched request bodies.
Fall back to content features and segment popularity while behavioural signal is absent, then blend towards collaborative scores as interactions accumulate. New items also need guaranteed exploration impressions, or the already-popular items take all the traffic and nothing can dislodge them.
A deadlock is a cycle in the wait-for graph, so the design goal is a graph that cannot contain one: stop holding two locks where you can, impose one total order on acquisition everywhere you cannot, and never call unknown code while holding a lock.
An order book is price levels in price order, each holding a time-ordered queue of orders, plus an ID index so a cancel is O(1); the hard parts are that cancels dominate the flow, the feed must be gap-checked and recoverable, and replay must be deterministic and allocation-free. It also connects matching engine to the point an interviewer is testing.
One exception. WhenAll's task faults with an AggregateException holding every failure, but await rethrows only the first inner exception, so recovering the rest means keeping the WhenAll task in a variable and reading its Exception property.
Write your stories in full, prepare the questions you will ask them, and stop early.
Give the dates, give the reason in one sentence, say what ended it, and stop. The interviewer is checking that your account matches the record and that you can discuss it without discomfort, not whether the reason was impressive.
Declare it, take the incident commander role explicitly, and mitigate before you diagnose - roll back or shed load first, understand later. Keep one channel as the record, publish updates on a fixed cadence, hand over rather than push through, and let the review produce owned actions instead of narrative.
Trace one real unit of work end to end, read the operational artefacts rather than the marketing ones, interview the people who handle exceptions, and publish a glossary inviting correction. Understanding shows as prediction and knowing why a rule exists; fluency only names things.
One-to-ones are the engineer's time for the things that do not fit anywhere else, not a status meeting. Feedback lands when it is specific, close to the event, about behaviour and its effect, and carries a request. Growth is owned by the engineer; the manager owns opportunity, honest signal and air cover.
Scope is the union of the track's sections, the rounds this employer type asks regardless of role, and the industry if you named one. Weights above 1.0 pull a section up the order; below 1.0 push it down. Nothing is excluded - a deprioritised section still appears when its best question is the most relevant thing left.