PayPal interview questions with answers (software engineer, India)
Last updated 2026-11-19
PayPal's India engineering teams work on payments, risk and platform services. Candidates commonly describe interviews that combine data structures and coding with backend topics specific to money movement: concurrency control, distributed transactions, idempotency and correctness of calculations, followed by behavioural questions.
The thirteen questions below are practice in that style, with worked answers. They are not past questions. For the online assessment, see PayPal coding round questions.
| Typical roles | Software Engineer (SDE), various levels (India) |
|---|---|
| Rounds, as commonly reported | Online assessment, coding interviews, design and behavioural rounds |
| Heavy topics | DSA, Java or Python, concurrency, distributed systems, SQL |
| Hiring news | See Hiring status below |
Rounds as commonly reported
Candidates commonly report an online assessment followed by several interviews: coding, a design or backend discussion and a behavioural conversation. Entry-level loops are lighter on design, but concurrency and database correctness still come up often.
Expect interviewers to push on failure cases: what if the network times out, the service restarts mid-transaction or two requests arrive together. Practise answering with the specific mechanism, such as a conditional update or an idempotency key.
What to prepare
- Data structures and algorithms in your main language.
- Concurrency control, database transactions and distributed-transaction patterns.
- API styles, caching and messaging: REST, gRPC, Redis and Kafka.
- SQL and money-handling correctness. See SQL window functions interview questions.
- Behavioural stories; see how to crack the HR interview and how to crack the technical interview.
Hiring status
For news that changes month to month, see PayPal hiring, which is updated from sourced reports. This page keeps only the facts below, each with its source.
Hiring may therefore be limited; check PayPal's own careers site for roles that are actually open.
- A September 2026 report says PayPal laid off about 220 employees in India (about 4%) across Chennai, Bengaluru and Hyderabad in technology, engineering, operations, payments and finance, as part of a multi-year transformation; early reports had said 600.
- The same report says PayPal reportedly plans to cut about 20% of its global workforce over several years.
PayPal interview questions with answers (software engineer, India): questions with worked answers
1 Two requests withdraw from the same wallet at the same moment. How do you stop the balance going negative?
hardA read-then-write in application code is a race: both requests read 100, both approve a 80 withdrawal. Make the check and update atomic in the database: a conditional update such as UPDATE wallet SET balance = balance - 80 WHERE id = ? AND balance >= 80, and check the affected-row count; or lock the row with SELECT ... FOR UPDATE inside a transaction; or use optimistic locking with a version column and retry. Pick the conditional update when you can, since it needs no explicit lock. Add an idempotency key so a retry is not applied twice.
2 Explain the saga pattern and why it is used instead of two-phase commit across services.
hardTwo-phase commit coordinates a prepare and commit across resources, but it blocks if the coordinator fails, holds locks and is rarely supported across independent services. A saga splits a business transaction into local transactions, each with a compensating action: reserve funds, then credit the recipient, and if the credit fails, refund the reservation. It can be orchestrated by a central coordinator or choreographed through events. Sagas give eventual consistency, so design compensations to be idempotent and expose intermediate states such as pending to users.
3 State the CAP theorem and say how it applies to a payments system.
mediumDuring a network partition a distributed system must choose between consistency (every read sees the latest write) and availability (every request gets a response). Partition tolerance is not optional in a real network. For ledger balances and settlement, choose consistency and accept refusing some requests during a partition. For things like product recommendations or notification history, availability with eventual consistency is acceptable. Also mention that CAP is about behaviour during partitions only, and latency trade-offs exist the rest of the time.
4 Compare REST, GraphQL and gRPC. Where would each fit in a payments platform?
mediumREST is simple, cacheable and widely understood, good for public partner APIs. GraphQL lets clients ask for exactly the fields they need, which helps mobile and web front ends with varied screens, but needs care with query cost and authorisation per field. gRPC uses Protocol Buffers over HTTP/2: compact, typed and fast, suited to internal service-to-service calls with streaming. Choose by audience: external partners, flexible clients or internal low-latency calls.
5 Design a peer-to-peer money transfer service at a high level.
hardStart with the API: POST /transfers with sender, receiver, amount, currency and an idempotency key. Validate, then in one database transaction debit the sender and credit the receiver and write ledger entries (double-entry, so each transfer has equal debits and credits). Use a risk check before committing, and publish an event after commit through an outbox table so notifications and analytics are reliable. Partition accounts by id for scale, use BigDecimal or integer minor units for money, and keep an audit trail. Discuss what happens on a timeout: the client retries with the same key.
6 SQL: find users who sent money to more than three distinct recipients in the last 7 days.
easyFilter to the last seven days, group by sender, count distinct recipients and keep groups above three. COUNT(DISTINCT) is the key; COUNT(*) would count repeat transfers to one person. Date arithmetic differs by database, so state the dialect. This is a typical simple risk signal.
sqlSELECT sender_id, COUNT(DISTINCT receiver_id) AS recipients FROM Transfers WHERE created_at >= CURRENT_DATE - INTERVAL '7 days' GROUP BY sender_id HAVING COUNT(DISTINCT receiver_id) > 3;7 Why must money never be stored as double, and what should you use in Java?
easyBinary floating point cannot represent most decimal fractions exactly: 0.1 + 0.2 is 0.30000000000000004. Errors accumulate across millions of transactions. Use BigDecimal with an explicit scale and RoundingMode, always constructed from a String or BigDecimal.valueOf, not from a double. Alternatively store integer minor units (paise or cents) in a long and keep the currency alongside. Compare BigDecimals with compareTo, not equals, which also compares scale.
javaimport java.math.*; class Money { public static void main(String[] args) { BigDecimal a = new BigDecimal("0.10"), b = new BigDecimal("0.20"); System.out.println(a.add(b)); // 0.30 System.out.println(new BigDecimal("10").divide(new BigDecimal("3"), 2, RoundingMode.HALF_EVEN)); // 3.33 } }8 What cache eviction policies does Redis support, and how do you choose one?
mediumCommon policies include allkeys-lru (evict the least recently used key from all keys), volatile-lru (only keys with a TTL), allkeys-lfu (least frequently used), random and noeviction, which returns errors when memory is full. For a pure cache choose allkeys-lru or allkeys-lfu; use noeviction when Redis holds data you cannot lose. Redis uses an approximated LRU sampling a few keys rather than an exact list. Set maxmemory and a TTL on cached entries so staleness is bounded.
9 How do Kafka consumer groups and partitions affect ordering and parallelism?
mediumA topic is split into partitions; ordering is guaranteed only within a partition. In a consumer group each partition is read by exactly one consumer, so parallelism is limited by the number of partitions. To keep events for one account in order, use the account id as the message key so they hash to the same partition. If a consumer fails, its partitions are reassigned, so processing must tolerate redelivery. Commit offsets after processing, not before, for at-least-once behaviour.
10 How do you run two independent service calls in parallel in Java and combine the results?
mediumCompletableFuture.supplyAsync runs each call on an executor, and thenCombine joins the two results when both finish. Pass a dedicated executor for blocking I/O instead of the common pool, add timeouts with orTimeout, and handle failures with exceptionally or handle. Total latency becomes the slower call rather than the sum. Mistake: calling get on the first future before starting the second, which makes the calls sequential.
javaimport java.util.concurrent.*; class Aggregator { static String load(ExecutorService ex) throws Exception { CompletableFuture<String> profile = CompletableFuture.supplyAsync(() -> "profile", ex); CompletableFuture<String> balance = CompletableFuture.supplyAsync(() -> "balance", ex); return profile.thenCombine(balance, (p, b) -> p + "+" + b).get(2, TimeUnit.SECONDS); } }11 Why PayPal, and why payments?
easyAnswer with what you find interesting in the problem: trust between strangers, fraud, scale and money moving across borders and currencies. Link it to something you have built or learned, such as handling concurrency or building an API. Because hiring conditions can change, show that you have read the role's own posting and say which team's work you would like to understand. Avoid commenting on company news you cannot back up.
12 Tell me about a time you worked with unclear requirements.
mediumDescribe how you clarified: asking the person who owned the requirement targeted questions, writing down assumptions, delivering a thin slice to get feedback quickly, and recording the decision. Give a concrete outcome and what you changed in how you start tasks. Interviewers want to see that you neither froze nor guessed silently.
13 Describe something you improved and how you measured the improvement.
mediumChoose an example with a number: a query that went from seconds to milliseconds after an index, a build that got shorter, a bug class eliminated by tests. Explain how you found the bottleneck (profiling, logs), what you changed, and how you measured before and after. State the number honestly; a modest, well-measured result beats a vague large one.
Frequently asked questions
Does PayPal ask system design?
For SDE roles with some experience, candidates commonly report design questions such as a transfer service. Early-career loops focus more on coding with some backend discussion.
Which topics are PayPal-specific?
Money correctness, such as BigDecimal and ledgers, idempotency, concurrency control and fraud signals tend to come up for payments companies.
Is hiring open right now?
Check PayPal's own careers site and the sourced notes in PayPal hiring, which include reports of layoffs in India in September 2026.
Which language should I use?
Use the one you are strongest in. Java and Python are common for backend roles.
Sources
Related guides
- PayPal coding round questions with solutions
- Visa interview questions with answers (software engineer, India)
- Mastercard interview questions with answers (software engineer)
- American Express interview questions with answers (engineering)
- System Design Interview Questions with Worked Designs
- SQL Window Functions Interview Questions
More in Company Interview Questions
- Paytm Coding Round Questions with Solutions
- Paytm Interview Questions and Answers
- Persistent Systems Coding Round Questions with Solutions
- Persistent Systems Interview Questions for Freshers
- PhonePe Coding Round Questions with Solutions
- PhonePe Interview Questions and Answers
- Postman Coding Round Questions With Solutions
- Postman Interview Questions and Answers
- Publicis Sapient Coding Round Questions
- Publicis Sapient Interview Questions with Answers
- PwC coding round questions
- PwC interview questions and answers