Currency Conversion (~3 parts): Given a list of exchange-rate strings in the format 'USD:CAD:1.3', implement Part 1 — direct conversion between two currencies; Part 2 — conversion through an intermediate currency when no direct rate exists (graph traversal); Part 3 — find the best conversion path across multiple hops, handling missing rates and bidirectionality.

Model currencies as nodes and rates as directed weighted edges, then apply BFS/DFS or Dijkstra for Part 2+. Solve Part 1 cleanly with a hash map in under 15 minutes to leave time for graph work. Validate inputs, handle unknown currencies explicitly, and name variables after the domain (fromCurrency, toAmount) — not generic letters. Graph modeling instinct, incremental problem extension, clean production-quality code. Stripe interviewers note this problem weeds people out fast because Part 3 separates candidates who understand transitive closure from those who only memorize LeetCode patterns.

Fraud / Chargeback Parser: Parse a stream of chargeback records (CSV format) to identify which transactions are disputed, filtering out malformed or invalid rows and extracting relevant metadata into a structured format.

Read the full prompt before writing a line — Stripe hides the real requirements inside lengthy business-context paragraphs. Parse defensively with explicit null/error branches. Output a typed struct, not raw strings. Interviewers expect a brief restatement of requirements before coding. Attention to spec, defensive parsing, and the ability to handle real-world messy data. The logic is intentionally straightforward; the test is whether the candidate reads carefully and handles edge cases without being told to.

Integration Round: Use the live Stripe API to build a working checkout or subscription flow for a described merchant scenario — e.g., implement a PaymentIntent creation, handle the confirmation response, and write an idempotency-safe retry wrapper around the charge call.

Read Stripe's developer docs before the interview so API object shapes (PaymentIntent, Customer, Subscription) are familiar. Start with the happy path, then layer in idempotency keys, error-code branching (card_declined vs. insufficient_funds), and logging. The interviewer has already seen bad merchant integrations in the wild — mirror good ones. Practical API fluency, idempotency awareness, and error-handling beyond the happy path. This round explicitly reflects real merchant mistakes Stripe has seen — interviewers are testing whether you'd write code they'd approve in a code review.

Design a distributed ledger for Stripe's payment processing that guarantees exactly-once semantics — every money movement is recorded once, debits always equal credits, and the system remains consistent across partial failures or network partitions.

Open with double-entry bookkeeping as the data model (every transaction has a debit entry and a credit entry in the same atomic write). Use event sourcing — immutable append-only records — rather than mutable balances. Address idempotency keys at the API layer, optimistic locking or database transactions at the storage layer, and an async reconciliation job to detect discrepancies. Discuss what happens when the upstream card network says 'yes' then sends a reversal. Financial invariant thinking ('total debits = total credits at every point in time'), exactly-once delivery mechanisms, and awareness that correctness trumps latency in payment systems. Blind reports identify this as the question that separates mid-level from senior/staff candidates.

Design Stripe's webhook delivery system: guarantee that every subscribed merchant receives each event at least once, handle retries with exponential backoff, minimize duplicate delivery, and allow merchants to verify event authenticity.

Model events as durable records in a queue (Kafka or Postgres-backed outbox pattern). Each delivery attempt is a separate record with status + retry count. Use HMAC-SHA256 signature headers for authenticity. Idempotent delivery: include an event ID and document that merchants must deduplicate. Discuss ordering guarantees (per-account FIFO vs. global), dead-letter queues for undeliverable events, and a dashboard for merchants to replay failures. At-least-once vs. exactly-once tradeoff articulation (you cannot guarantee exactly-once over a network — this single fact drives all operational decisions), outbox/transactional messaging pattern, HMAC for webhook security, and understanding of how Stripe's actual webhook product works.

Design Stripe Radar: a real-time fraud detection system that scores every payment transaction as it occurs — at Stripe's scale — and blocks or flags high-risk transactions before authorization is sent to the card network.

Decompose into: feature extraction (velocity checks per card/IP/merchant in the last N minutes using Redis sorted sets), a model serving layer (pre-trained ML model loaded in-memory for sub-100ms p99 latency), rules engine (merchant-configurable thresholds), and feedback loop (chargebacks feed labeled data back to retraining). Discuss false-positive vs. false-negative trade-offs: a false positive blocks a legitimate customer, a false negative lets fraud through — Stripe lets merchants tune this. Latency budget awareness (must complete before the network times out, ~100–200ms total), velocity-check data structures (sliding window counters), and the business trade-off framing. ML candidates are also probed on feature engineering and model drift.

Tell me about a time you disagreed with another engineer on a technical decision. How did you handle it, and what was the outcome?

Use a concrete engineering disagreement (architecture, API design, rollout approach) — not an interpersonal conflict. Show you raised the concern with data or a prototype, not just an opinion. Describe how you listened to their reasoning, what changed your mind or didn't, and how you reached alignment without needing escalation. End with what you'd do differently. Technical conviction paired with intellectual humility. Stripe explicitly tests whether candidates can advocate for a position without ego and change their mind when shown better data. Candidates who 'always deferred' or 'always won' both raise flags.

Tell me about a production incident or serious bug you owned end-to-end — from discovery through root cause analysis to prevention.

Walk through detection (how did you find out?), immediate mitigation (what stopped the bleeding?), root cause (what actually went wrong?), and post-mortem (what changed so it can't recur?). Quantify customer impact if possible. Show you wrote the post-mortem and drove the follow-up items, not just fixed the immediate issue. Ownership culture (Stripe values engineers who treat reliability as personal responsibility), systematic debugging under pressure, and blameless post-mortem thinking. The interviewer wants evidence you learn from failures and share that learning.

Describe a situation where requirements changed significantly mid-project. How did you handle the ambiguity and re-plan?

Pick a real case where the change was material — not a one-line spec update. Describe how you quickly assessed impact (what's still valid, what must change, what's now waste), re-prioritized with stakeholders, and communicated the new plan without blaming anyone for the pivot. Show you kept the team unblocked during the transition. Adaptability and judgment under ambiguity — Stripe operates fast and requirements genuinely shift. The interviewer is checking whether you freeze, escalate everything, or exercise independent judgment to keep moving.

Stripe Interview Questions

Stripe's SWE loop (2024–2026) runs 5–6 rounds: online assessment, technical phone screen, then an onsite with distinct Coding, Integration, Bug Bash, System Design, and Behavioral sessions. The process is deliberately practical — problems mirror real Stripe infrastructure challenges, and interviewers score production-readiness (clear names, error handling, edge cases) as heavily as correctness.

12 real questions across 3 rounds

Reported via LeetCode Discuss (Stripe company tag, Sep 2024 reject thread; 2026 new grad OA thread), bigtechexperts.com Stripe question bank (page titled 'Q1 - Stripe Currency Conversion Problem (~3 Parts)'), interviewing.io Stripe guide, Medium candidate write-up by @diyaag2020 (2025–2026) · Teamblind 'Stripe Onsite Interview – Bug Squash' thread, interviewing.io Stripe guide, blog.rampatra.com first-person Stripe Dublin interview write-up, Glassdoor Stripe SWE onsite reports, Prepfully Stripe SWE exhaustive guide

3 more coding questions Stripe actually asks, each with how to approach it and exactly what the interviewer is evaluating. Unlock the full set across every round.

Continue with Google

Or run a Stripe mock with the AI interviewer. It asks questions like these, follows up on your answers, and tells you exactly what to fix. The only score that means anything is the one on your real answers.

Continue with Google

Keep prepping

All company interviews