Jobra
๐ŸžStep 1 of Jobra:AI Debugging Rounds
DSA+AIPreparing you for both.
JOBRA ยท AI DEBUG ROUND (PREVIEW)sum.cpp
// buggy snippet
int total(vector<int>& a, int n) {
ย ย ย ย int sum = 0;
ย ย ย ย for (int i = 0; i <= n; i++) { // ๐Ÿžย ย ย ย ย ย ย ย sum += a[i];
ย ย ย ย }
ย ย ย ย return sum;
}
A real codebase. A real bug. You + AI.

Jobra drops you into a broken repo. Find it, fix it, ship it. AI helps you debug like on the job.

Prompt:Correct the loop condition
Tap 'Let Jobra fix it' below to simulate AI debugging...

Stripe Software Engineer Interview Questions & Rounds

19 min read

Stripe software engineer interview guide: the practical multi-part OA, team screens and onsite rounds on parsing, integration and debugging.

In this article
On this page

How to crack Stripe: The Software Engineer interview, from real experiences

Last reviewed November 17, 2026 ยท 19 min read ยท For candidates targeting a general / experienced Software Engineer role (mid-level focus), mixed/global geography

Prepared by an independent authoring agent for Codinzhub, not by or for Stripe. Evidence is the assigned Codinzhub/PracHub Software Engineer reports, one 2022 offer report, and public candidate accounts. Reports are individual experiences; your recruiter can confirm the current format.

The 30-second version. Candidates describe an application, a short recruiter or HR screen, an online assessment (OA โ€” a timed coding test), one or more technical phone or team screens, and a virtual onsite (VO โ€” the final loop of several one-hour rounds). The OA and screens are unusually practical: one long, multi-part task where each part builds on your earlier code, not a single algorithm puzzle. VO rounds reported include programming, an AI-assisted coding exercise, debugging, integration against an unfamiliar API, design, and an experience conversation. A 2022 report reached an offer after a five-part onsite. The most useful habit: get part one working and readable, then extend it while narrating what you do.

Important things to know before the interview

This is implementation, not a LeetCode drill. Across the reviewed accounts, OA and coding rounds are long, stateful tasks with progressive parts: deployment windows, a data-centre router, a connection load balancer, reconciliation, form validation. Interviewers appear to value correct output and clean structure. (Multiple candidate reports.)

The multi-part format is the main challenge. Part two adds a rule to part one, part three adds another, and old tests must still pass. One candidate wrote nearly 500 lines of Python across three parts and still felt rushed. Keep helpers small and know your editor. (Candidate-reported; 500 lines is one account.)

The AI round genuinely permits AI โ€” for that round. HackerRank (an online coding platform) provides an AI chat you prompt to read the brief, write code and add tests. Feedback differs: one candidate was told the AI's output was enough, another that they had not explained their thinking. Verify the rule for your round; do not generalise. (Candidate-reported.)

"Stripe discourages C/C++" is a research lead, not a rule. One 2025 India intern report says a prep doc ranked C/C++ below other languages; Stripe's job posts list commonly used languages but no interview ban. Other candidates prefer Python for speed or use Java. Ask the recruiter which languages your round supports. (Scope: one intern lead plus official posting context.)

What this guide covers

This guide targets one role and level: Software Engineer, general/experienced hiring with a mid-level focus. It excludes interns, new grads, senior and staff loops, and other disciplines. Where an external account is new-grad or senior, the table says so.

The reviewed Codinzhub reports were published from August to October 2026, with one offer report from November 2022. Geography is mixed: US work authorization, New York, fully remote roles and India/Bangalore appear. External research adds remote-US and New York accounts. Because the sample is small, the guide says "reported examples" rather than percentages. Most reports do not state a college, so experience rows say Not stated. Your recruiter is the authority on your sequence.

Round-by-round breakdown

This order comes from candidate reports, not a published official sequence; steps appear to vary by team, level and location.

Stage 1: Application, where candidates applied

No official channel list was verified, but accounts describe:

ChannelWhat candidates reported
Direct online applicationApplied on Stripe's jobs site; an OA sometimes arrived the same day ("mass send-out" in one account), sometimes after a long wait
Employee referralA referral produced an OA "right away" for a mid-level applicant; another got one two days later
Recruiter outreachRecruiter emails and LinkedIn Inmail, including a cold email for a remote-US L4 role
Campus / off-campusOne 2025 India intern portal application led to an OA two months later โ€” intern scope, not this role

Stage 2: Shortlisting, how candidates moved forward

Usually a recruiter or HR conversation. One US report lists ten questions from a 20-minute screen; others describe location, compensation, past work and motivation. This is what one candidate was asked, not a script.

At a glance: 20 minutes in one report, about 25โ€“30 in another; ten questions in the detailed account; topics of background, motivation, logistics, level and AI usage.

Question summarySkill or patternReported follow-upLevel / scopeCodinzhub link
"Tell me a bit of background on yourself."Self-introductionNot reportedUS, 2026HR screen
"How have you used AI in your career?"AI experienceNot reportedUS, 2026HR screen
"Are you authorized to work for any US-based employer?"Work authorizationNot reportedUS scopeHR screen
"How would you deal with working with a company you weren't aligned with?"Values / situationalNot reportedUS, 2026HR screen
"Why is Stripe of interest to you?"MotivationLocation, work and pay discussedUS / remoteHR screen
"Which language would you use for the next step?"Language preferenceNot reportedUS, 2026HR screen
"Where else are you interviewing, and how far along?"Job-search statusNot reportedUS, 2026HR screen
"Are you open to other locations?"Location preferenceNot reportedUS, 2026HR screen
"What level are you looking at โ€” mid or senior?"Level alignmentNot reportedUS, 2026HR screen
"Does this compensation range align with you?"CompensationNot reportedUS / remoteHR screen

Stage 3: Online assessment (OA)

How it works. A timed take-home coding test, usually on HackerRank. Reports describe one long, multi-part problem, not several short ones; later parts are often hidden and reuse your earlier code.

At a glance: 45โ€“90 minutes (60 most common); one task of two to five parts; topics of intervals, time zones, sorting, parsing, state machines and basic geometry.

Question summarySkill or patternReported follow-upLevel / scopeCodinzhub link
Given allowed and frozen time ranges in a week, return deployable windows = allowed minus freeze.Interval merge, sweep line (sort endpoints, track active ranges)Part 2 adds rulesMid-level, referral, 2026Two-part scheduler OA
Add time-zone offsets, a current UTC time, lead time, minimum duration and limit k; return at most k UTC windows.Time zones, week wrap, greedy splitExtension; 10/11 cases in one report2026UTC deployment windows
Implement REGISTER region lat lon capacity and SET_HEALTHZ region state with validation.Command parsing, validationDistance and routing follow2026Data-centre OA
Implement DISTANCE lat1 lon1 lat2 lon2 via the Haversine formula (great-circle distance), rounded.GeometryExtension2026Data-centre OA
Implement ROUTE lat lon to the nearest healthy region; break ties alphabetically; output NONE 0 if none.Sorting, tie-breaksFinal part2026OA and tree VO
Simulate a Jupyter Server connection router: CONNECT assigns a target server.Stateful simulation, round-robinFour further parts2026Load balancer OA
A duplicate CONNECT on the same id keeps the original target and changes no state.id โ†’ target mapExtension2026Load balancer OA
DISCONNECT closes a connection, frees the slot and invalidates the id.Resource cleanupExtension2026Load balancer OA
Enforce maxConnectionsPerTarget; skip full servers, ignore if all full.Capacity checksExtension2026Load balancer OA
SHUTDOWN targetIndex takes a target offline and evicts its connections in order.Batch evictionFinal part2026Load balancer OA

Stage 4: Technical screen, team screen and VO programming

How it works. A 45โ€“60 minute coding conversation on HackerRank or in your own editor with screen sharing ("pair programming"). Requirements arrive in stages; interviewers watch output correctness and corner cases such as missing keys, invalid numbers and sorting.

At a glance: 45โ€“60 minutes; one multi-part task, sometimes two; strings, parsing, weighted matching, class design, forms, CSV and trees.

Question summarySkill or patternReported follow-upLevel / scopeCodinzhub link
Given user records with name, email, company and field weights, find records matching a target when weighted similarity meets a threshold.Weighted scoring, similarityOne-hop then full connected componentPhone screen, 2026Fuzzy matching
Include users matched indirectly through another user (one hop).Graph traversalExtension2026Fuzzy matching
Return the whole linked group regardless of hop count.Connected componentsExtension2026Fuzzy matching
Three-part "calculate cost" problem; parts two and three add rules while earlier cases must still pass.Incremental implementation, regression safetyCorner cases watched; nearly 500 lines of Python in one accountRemote, 2026Three-part screen
Given bank/account/date/balance rows, return the latest balance for a bank and account.Sorting, selectionOutput-format ambiguity2026Blank editor screen
Compare two monthly CSV tables and mark matching bank-account balances.Two-table comparisonThird part not seen2026Blank editor screen
Build a country selector, passport input and validate button; validate a number against that country's format (L letters, D digits, A alphanumeric); then wire errors, resets and accessibility.Form layout, pattern validation, state wiringTest scenarios; production readiness2026Passport form
Implement AccountScheduler.is_available(id, t) from account ids and lock-until timestamps.Class designacquire, then LRUVO coding, 2026Fuzzy matching + VO
Follow-ups: add acquire(id, duration) to extend a lock, then auto-pick the least recently used (LRU) available account when none is named.State mutation, LRU policyReported follow-ups to is_available2026Fuzzy matching + VO
Parse an input format and run simple operations with basic data structures (60-min pair programming).Parsing, fundamentalsFinish over optimalitySan Francisco, 2022, offeredPractical coding
Render nested task records as a tree with โ”œโ”€ and โ””โ”€ connectors.Tree building, formattingLong statementVO, 2026Tree-rendering VO

An assigned report names a three-part KYC validation task (company name, URL, abbreviation, description) with no usable detail: KYC report.

Stage 5: The virtual onsite (VO)

Round counts of three, four and five appear in reports. Common ingredients: programming, AI programming, debugging, integration, design and an experience conversation. One report says the final block was described as three to four hours and could be split over two days.

Stage 5.1 AI programming exercise

HackerRank with an AI chat: you get transaction records and ordered rules and decide accept or block, starting with string matching and growing more complex. One candidate had about 30 minutes of coding time. What is scored is unclear from reports.

At a glance: about 30โ€“60 minutes; one multi-part rule parser; parsing, Boolean logic, tests.

Question summarySkill or patternReported follow-upLevel / scopeCodinzhub link
String-equality rules on transaction fields; rules applied in order, first match wins, default accept.Ordered rule parsingBoolean operators next2026Reconciliation and AI exercise
Add AND/OR conditions, including a merchant name containing "and" or "or".Boolean logic, tokenisationNested conditions2026Reconciliation and AI exercise
Support complex combined conditions with parentheses.Recursive parsingMore parts existedL3 onsite, 2026Five onsite rounds

Only three distinct questions were verified for this round โ€” the guide's main evidence shortfall.

Stage 5.2 Debug / Bug Squash

You get an unfamiliar codebase with failing tests: reproduce, read, hypothesise, fix and explain the root cause. Set up your IDE and plugins beforehand โ€” one candidate lost half an hour to setup. A 2022 report describes downloading a codebase revision to find a bug.

At a glance: 45โ€“75 minutes; one or more repository bugs; HTTP redirects, templates, YAML/CSV, currency precision.

Question summarySkill or patternReported follow-upLevel / scopeCodinzhub link
In Python requests, make three tests pass so a POST resends its body after an HTTP 307 redirect, including a partly read stream.Stream cursors, HTTPExplain root cause2026307 bug
Debug the Mako template engine.Reading unfamiliar codeTime pressureL3, 2026Five onsite rounds
Fix SnakeYAML so flag: on parses.YAML parsingSecond bug followsJava track, 2026AI, debug and integration
Fix SnakeYAML CSV parsing so quotation marks survive.CSV parsingโ€”Java track, 2026AI, debug and integration
Find a floating-point precision bug in currency conversion at certain exchange rates.Numeric precisionOne of five planted bugsNew-grad report, 2026Not on Codinzhub; external candidate report

Stage 5.3 Integration

You extend code that calls an API or library you may not know, with docs available. Reports stress reading docs quickly and not breaking earlier behaviour.

At a glance: 45โ€“60 minutes; one multi-part task; JSON/GeoJSON, HTTP, maps, owner lookup.

Question summarySkill or patternReported follow-upLevel / scopeCodinzhub link
Parse a bike-ride GeoJSON file and print the first ten coordinates.JSON parsingParts 2 and 32026BikeMap detail
POST map parameters to a static-map API; save the PNG response.HTTP POST, binary dataAdd markers and paths2026BikeMap detail
Add start/end markers and alternating coloured paths across the coordinates.Data transformationFinal part2026BikeMap detail
Match payments to invoices using an id in the payment memo.Parsing, map lookupAmount and tolerance rules2026Reconciliation
Add exact-amount matching, choosing the earliest-due invoice on ties.Tie-breakingForgiveness range2026Reconciliation
Add a tolerance ("forgiveness") range and discuss priority when exact and range match.Edge casesInterviewer asked for more cases2026Reconciliation
Use JGit to compare two branches; from a CSV of owners, return who owns most changed files.Library integration, aggregationDocs unavailableJava track, 2026Integration questions

Stage 5.4 System design

An open-ended discussion. Reports range from playful prompts to payments infrastructure, pushing on consistency, idempotency (repeating an operation has the same effect as doing it once), retries and justification of choices.

Question summarySkill or patternReported follow-upLevel / scopeCodinzhub link
Design a superhero dispatch system: an accident triggers a dispatch, a nearby hero accepts.Matching, dispatchWeak tech justificationL3, 2026Five onsite rounds
Design a highly idempotent payout pipeline.ACID, consistency, retries, outboxDeep edge casesL4 backend, remote US, 2026Not on Codinzhub; candidate report
Design an in-memory rate limiter with sliding windows, distributed enforcement and per-customer overrides.Rate limiting, concurrencyToken bucket vs leaky bucketBackend, NYC, 2026Not on Codinzhub; candidate report
Design a payments platform around consistency, latency, idempotency and failure handling.Payments fundamentalsโ€”L3, community-sourcedNot on Codinzhub; community report
Design a financial ledger plus Bikemap-style external integration with rate limits and SLAs.Ledger, adaptersโ€”Community practice, 2026Not on Codinzhub; community source

Stage 5.5 Experience, goals and behavioral

Reported as "Behaviors & goals" or a manager conversation about projects and customer communication. A 2022 onsite includes an "Experience & Goals" round and a career-history talk.

Question summarySkill or patternReported follow-upLevel / scopeCodinzhub link
Describe a time you built an MVP and then had to keep improving it.Ownership, iterationCandidate under-prepared2026AI coding and debugger
Questions about communicating with customers.Customer empathyPart of manager roundL3, 2026Five onsite rounds
A behavioral discussion about adaptability.Adaptabilityโ€”New York, 2026Practical coding
Present a former technical project, then answer questions such as what you would do differently.Project depth, reflectionQ&A2022 offer and 2026Practical coding

Final decision

No official role-specific decision policy was verified. Reports mention rejection emails or calls, feedback such as "debug and integration didn't meet the bar," a one-year reapply cooldown in one case, and no offer after rounds the candidate felt went well. Treat timing and feedback as individual examples.

Stripe experiences shared on Codinzhub

College or backgroundRole and reported outcomeWhat was askedExperience
Not statedSoftware Engineer, 2022 โ€” OfferedRecruiter screen, parsing phone screen, five-part onsiteOffer report
Not statedSoftware Engineer, 2026 โ€” no offerHR screen, deployment OA, coding, debug, integrationReconciliation and 307 bug
Not statedSoftware Engineer, 2026 โ€” advanced toward VOHR screen, three-part cost calculation, deployment OAThree-part screen
Not statedSoftware Engineer, 2026 โ€” not reportedData-centre OA, tree-rendering VOOA and tree VO
Not statedSoftware Engineer, 2026 โ€” no offerAI coding, reconciliation, requests debug, MVP behavioralDebugger failure

The Codinzhub next-round playbook

These are editorial recommendations tied to reported demands; none is claimed to have caused an offer.

For the OA and coding screens

  1. Practise one long problem in a single sitting; rehearse adding rules without rewriting.
  2. Keep fast helpers ready: interval merge, deterministic sort with tie-breaks, CSV/JSON parsing, "key exists" checks.
  3. Know your IDE cold. Candidates credit it for quick typo fixes; one lost time to environment setup.
  4. Dry-run by hand with an empty input, a duplicate and a boundary value before submitting.
  5. Narrate your decisions; candidates were told to think out loud.

For the VO

  1. For debugging, break a test in a small open-source repo and practise reproducing, isolating and explaining it. Confirm your debugger starts.
  2. For integration, skim one library's docs and write a tiny request/response exercise.
  3. For AI coding, rehearse a clear brief to the assistant, then review and test its output.
  4. For design, state assumptions and justify each technology choice.
  5. For behavioral rounds, prepare four short true stories.

Variations within this role

Location and route. Remote-US accounts describe a recruiter screen, a technical screen extending an API framework with unit tests, and a four-round VO. New York reports mention practical coding plus design, integration, behavioral, project presentation and hiring-manager talks. India reports include OAs sent soon after application or referral and an intern-only language note โ€” intern scope and only a lead. Do not assume these sequences travel between locations.

Year. The 2022 offer report has a classic five-part onsite with debugging and design. Reports from 2025โ€“2026 add the AI coding round and more practical OAs. AI assistance appears in newer accounts; confirm it with your recruiter.

Round names. "Team screen," "technical phone screen," "programming" and "VO" are used inconsistently. One candidate's phone-screen question became another's VO question. Focus on task shape, not labels.

Do's and don'ts

DoDon'tWhy it matters
Confirm the language and tools per roundDon't assume C/C++ is banned or Python mandatoryLanguage claims are preferences or one intern lead, not verified rules
Keep part one clean and smallDon't hard-code for the sampleLater parts add rules and old cases must pass
Test missing keys, invalid numbers, ties, empty input, week wrapDon't stop after the example worksInterviewers in the reports watched for corner cases
Use AI in the AI round and explain its outputDon't rely on it where it isn't permittedPermission is round-specific
Check debugger, IDE and plugins before the callDon't fix your environment during the roundOne candidate lost the debug round this way
Prepare honest STAR stories (Situation, Task, Action, Result)Don't invent achievementsFeedback cited thin, execution-only answers

Most-asked topics

Frequencies below are "reported in the reviewed accounts," not percentages.

TopicWhat it meansPriorityReported examplesCodinzhub practice
Multi-part implementationSolve a task, then extend it as rules arriveRecurring in almost every reportDeployment windows, cost calculationMerge Intervals
Intervals and schedulingMerge allowed ranges, subtract frozen onesRecurring in OAsDeployment schedulerInsert Interval
Parsing and validationRead structured text into dataRecurringCSV balances, passports, YAMLFirst Unique Character
Stateful classes and cachesObjects whose state changes over timeRecurringAccountScheduler, connection router, LRULRU Cache
Graphs and componentsLink directly or indirectly related recordsReported with follow-upsFuzzy matchingAccounts Merge, Number of Islands
TreesBuild parent/child links from flat recordsVO and design tasksTree renderingLevel Order Traversal
Basic geometryDistance from coordinatesTwo-part OAHaversine DISTANCEHaversine formula
Conditional evaluationOrdered rules with Boolean logicNewer AI roundTransaction rule parserValid Parentheses
Payments fundamentalsIdempotency, consistency, retriesDesign roundsPayout pipeline, ledger, precision bugDesign HashMap

The question patterns that keep coming back

Progressive constraints. The dominant pattern: solve something, then reveal a rule that must not break old tests. Look for "parts" or hidden later stages. Keep validation and core logic in small helpers; candidates repeatedly mention running out of time.

State machines and simulated systems. Deployment windows, connection routers, account schedulers and payment states ask you to model changing state. Draw states and transitions first. The window reports highlight UTC conversion wrapping past the week end, splitting a continuous interval in two; one candidate expanded the timeline to three weeks.

Records, matching and tie-breaks. Invoice reconciliation, user matching and routing all choose a "best" record. Expect tie-break rules such as earliest due date, alphabetical order or highest similarity. Write the comparison explicitly and test the tie.

Real-library debugging and integration. These rounds use real repositories. The skill is reading failing tests, forming a hypothesis, using your debugger and explaining the cause. One report traces a POST body re-sent after an HTTP 307 redirect with the stream cursor at the end, so the request has a Content-Length but no bytes to send.

Preparing for behavioral and experience rounds

The reported style is conversational and project-led, not a fixed competency script. Prepare real stories with a situation, your action, the result and what you would do differently โ€” STAR is a memory aid, not a script. Cover: iterating on an MVP, disagreement or alignment, customer communication, adaptability, and a call made with incomplete information. One L3 report describes many project and customer questions; another says answers leaned too much toward execution. Speak about impact and learning, not just tasks.

Practice problems that match the experiences

These are practice analogues, not problems confirmed to have been asked. Custom Stripe tasks have no official difficulty rating here.

#Practice problemPatternRelated reported variant
1Merge IntervalsInterval mergingDeployment windows
2Insert IntervalInterval insertionAdding a rule to a range set
3LRU CacheCache designAccountScheduler LRU follow-up
4Accounts MergeComponentsIndirect user matching
5Number of IslandsGraph traversalFinding a linked group
6Valid ParenthesesStack, nestingRule conditions with parentheses
7Design HashMapHashingMemo matching, owner lookup
8Top K Frequent ElementsHeap / countingReturning at most k windows
9Level Order TraversalTree traversalRendering nested records
10First Unique CharacterString scanningParsing basics
11Haversine formulaGeometryDISTANCE command

How to handle a question you have never seen

  1. Restate the problem. Say back inputs, outputs and constraints; ask about empty or duplicate data.
  2. Work one example by hand. A dry run checks your logic step by step on a small sample.
  3. State a simple first approach. Get part one working even if not optimal; candidates report correctness matters more.
  4. Find the bottleneck. Ask which piece breaks when a new rule arrives; isolate it in a helper.
  5. Improve and extend. Add the requirement while keeping old tests passing.
  6. Test deliberately. Boundaries, ties, invalid data and "no match".
  7. Explain while you work, especially when reviewing AI-generated code.

Why candidates were rejected

  • Explicit recruiter feedback (one report): debug and integration "didn't meet the bar," though overall performance was called positive; a one-year cooldown was mentioned.
  • Explicit interviewer feedback (two reports): weak technology justification in design; not enough explanation of thinking in the AI round.
  • Candidate interpretation (not verified): one weak round, a confusing prompt, an unfamiliar interviewer, or a debugger that would not start. These are the candidates' readings, not confirmed causes.
  • Unknown: many reports give no reason. A delay or incomplete solution alone is not evidence of why.

If you remember only five things

  1. Expect one long, multi-part, practical task rather than a short algorithm puzzle.
  2. Keep part one clean and extensible; retest old cases after every change.
  3. Confirm the language and AI policy for your round; the C/C++ claim is only a lead.
  4. For debugging, verify your environment first, then reproduce, isolate and explain the root cause.
  5. Prepare short, honest stories about iteration, customers and incomplete-information decisions.

FAQ

Is Stripe's Software Engineer OA LeetCode-style?

Reported accounts say no: a long, implementation-heavy, multi-part task on intervals, parsing, state machines and geometry. One report suggests LeetCode 56 and 1229 as related practice, not exact matches.

How long is the process?

OAs of 45โ€“90 minutes, screens of 45โ€“60, and VO blocks of three to five one-hour rounds are reported. One final block was described as three to four hours, optionally split over two days. Two candidates report about a month of scheduling delays. None of this is a promise.

Is system design asked?

Yes in some reports: a 2022 offer report, a 2026 L3 design prompt and a remote-US L4 payout pipeline. One senior account reports no design round. Ask your recruiter.

Can I use AI tools?

The AI programming round explicitly provides an AI chat in HackerRank. Other rounds are less clear: one candidate says Google was allowed in a screen, and no universal rule was verified. Confirm per round.

Which language should I choose?

What you can write fast and explain clearly. Candidates report Python for speed and Java in Java-track rounds; one intern lead suggests C/C++ was ranked lower, while Stripe's posts list Ruby, Java, JavaScript, Go and Scala. No verified ban exists.

Does the process differ by location?

Examples differ: remote-US accounts emphasise API-framework extension and clean code; New York reports include design and project presentation; India reports are OA-heavy and partly intern. These are examples, not one global sequence.

What behavioral questions should I expect?

Themes reported: background, motivation, alignment with a company, MVP iteration, customer communication, adaptability and project deep dives. Prepare true stories, not scripts.

About this guide

Researched by an independent authoring agent for Codinzhub. It reviewed the assigned Codinzhub Software Engineer experiences, one 2022 offer record from company practice data, and public candidate accounts found through Reddit and other sources on November 17, 2026. Questions were paraphrased faithfully; none was invented, and practice analogues are labelled. External research URLs and the claim-level evidence ledger are kept in private research notes. Corrections are welcome through Codinzhub's contribution channels. Processes change and teams differ; your recruiter can confirm the current format.

Other company and role guides

These are related Codinzhub guides for nearby roles and companies. They are separate guides, not part of this role's evidence. Each link was checked against the published site.

Editorial completion checklist

  • Separate guide for the exact role and level
  • Reddit research completed and recorded privately
  • External research expanded when internal experiences were limited
  • At least five actual sourced questions per round โ€” AI programming has 3 (shortfall 2)
  • Important company-specific findings checked and highlighted
  • Application and shortlisting sections included
  • Do's and don'ts included
  • Most-asked topics and recurring patterns included
  • Matching practice problems included โ€” analogue disclaimer present
  • Relevant practice links verified โ€” 10 LeetCode and 1 GeeksforGeeks pages fetched; premium-only and 404 links removed
  • SEO metadata and beginner readability reviewed โ€” one H1, terms explained
  • Contextual links and related guides verified โ€” related-guide URLs checked against the published site
  • External research/resource links excluded from the public article
  • No calendar-based study plan included
  • Remaining gaps documented and draft/completion status accurate
RoundDistinct reported questionsMinimum five metRemaining gap
Stage 2 โ€” Recruiter / HR screen10YesNone
Stage 3 โ€” Online assessment10YesConservative base-problem count is 3; 7 rows are progressive parts
Stage 4 โ€” Technical / VO programming12YesConservative base-problem count is 8; KYC noted separately with no detail
Stage 5.1 โ€” AI programming exercise3No2 more distinct questions needed; all accounts describe one rule-parser task
Stage 5.2 โ€” Debug / Bug Squash5Yes (caveated)1 row new-grad; 2 rows from one report; Mako named only
Stage 5.3 โ€” Integration7Yes2 senior-scoped and 1 Java-track row
Stage 5.4 โ€” System design5Yes (caveated)2 of 5 rows community-sourced
Stage 5.5 โ€” Experience / behavioral4 table rows (+ screen questions)Yes (caveated)Detailed first-party behavioral bank is thin

Status: DRAFT. The AI programming round falls below the five-question minimum, and several other rounds clear the count only with caveated evidence.

Keep preparing on Codinzhub

Search the question bank

The questions from each round above, with practice links for similar problems. Log in to search every question we collected for this role, including questions shared on Codinzhub.

Showing 20 of 83 questions

  • "Tell me a bit of background on yourself."

    Level: Not reported.

    Self-introductionCodingRound not stated
  • "How have you used AI in your career?"

    Level: Not reported.

    AI experienceCodingRound not stated
  • "Are you authorized to work for any US-based employer?"

    Level: Not reported.

    Work authorizationCodingRound not stated
  • "How would you deal with working with a company you weren't aligned with?"

    Level: Not reported.

    Values / situationalBehavioralRound not stated
  • "Why is Stripe of interest to you?"

    Level: Location, work and pay discussed.

    MotivationCodingRound not stated
  • "Which language would you use for the next step?"

    Level: Not reported.

    Language preferenceCodingRound not stated
  • "Where else are you interviewing, and how far along?"

    Level: Not reported.

    Job-search statusCodingRound not stated
  • "Are you open to other locations?"

    Level: Not reported.

    Location preferenceCodingRound not stated
  • "What level are you looking at โ€” mid or senior?"

    Level: Not reported.

    Level alignmentCodingRound not stated
  • "Does this compensation range align with you?"

    Level: Not reported.

    CompensationCodingRound not stated
  • Given allowed and frozen time ranges in a week, return deployable windows = allowed minus freeze.

    Level: Part 2 adds rules.

    Interval merge, sweep line (sort endpoints, track active ranges)CodingOnline assessment
  • Add time-zone offsets, a current UTC time, lead time, minimum duration and limit k; return at most k UTC windows.

    Level: Extension; 10/11 cases in one report.

    Time zones, week wrap, greedy splitCodingOnline assessment
  • Implement REGISTER region lat lon capacity and SET_HEALTHZ region state with validation.

    Level: Distance and routing follow.

    Command parsing, validationCodingOnline assessment
  • Implement DISTANCE lat1 lon1 lat2 lon2 via the Haversine formula (great-circle distance), rounded.

    Level: Extension.

    GeometryCodingOnline assessment
  • Implement ROUTE lat lon to the nearest healthy region; break ties alphabetically; output NONE 0 if none.

    Level: Final part.

    Sorting, tie-breaksCodingOnline assessment
  • Simulate a Jupyter Server connection router: CONNECT assigns a target server.

    Level: Four further parts.

    Stateful simulation, round-robinCodingOnline assessment
  • A duplicate CONNECT on the same id keeps the original target and changes no state.

    Level: Extension.

    id โ†’ target mapCodingOnline assessment
  • DISCONNECT closes a connection, frees the slot and invalidates the id.

    Level: Extension.

    Resource cleanupCodingOnline assessment
  • Enforce maxConnectionsPerTarget; skip full servers, ignore if all full.

    Level: Extension.

    Capacity checksCodingOnline assessment
  • SHUTDOWN targetIndex takes a target offline and evicts its connections in order.

    Level: Final part.

    Batch evictionCodingOnline assessment
Page 1 of 5
Ready for the next topic?

Keep track of what youโ€™ve learned.