Product Company · Bengaluru
Flipkart's interview process focuses on DSA and system design for tech roles, with culture-fit rounds for product and business roles.
Process: Online Coding → Technical × 2 → Hiring Manager
What is the Flipkart interview process for software engineering roles?
Tip: Flipkart's process: online coding round (2 DSA problems, 90 min) → 2 technical rounds (DSA + CS fundamentals + system design basics) → hiring manager round (design + culture fit). Focus on clean, correct code — Flipkart reviewers prioritise correctness over speed.
Design a product search and recommendation system for an e-commerce platform at Flipkart's scale.
Tip: Search: Elasticsearch for full-text + filters (category, brand, price). Recommendation: collaborative filtering (user-item matrix), content-based (product embeddings), and real-time click-stream (Kafka → stream processing → feature store). Personalisation: A/B tested models per user segment.
Given a list of transactions, find all accounts that performed more than N transactions in one day.
Tip: Group by account_id + date using a HashMap<(account, date), count>. Flipkart data engineering rounds often involve aggregation problems. If using SQL: GROUP BY account_id, DATE(timestamp) HAVING COUNT(*) > N.
Explain how consistent hashing works and why it matters for distributed caching.
Tip: Consistent hashing: place both cache nodes and keys on a ring (0 to 2³²). A key is served by the next clockwise node. Adding/removing a node only remaps ~K/N keys (K = keys, N = nodes) instead of all K keys. Virtual nodes ensure even distribution. Used in Flipkart's CDN and session cache layers.
What is a B+ tree and why do databases use it over a binary search tree for indexing?
Tip: B+ tree: balanced m-way tree, all data in leaf nodes (linked for range scans), internal nodes store only keys. Advantages over BST: high branching factor reduces tree height (fewer disk I/Os), sequential scans via leaf linked list. MySQL InnoDB uses B+ trees for all indexes.
How would you implement a distributed rate limiter that works across multiple API servers?
Tip: Centralised Redis with atomic Lua scripts (INCR + EXPIRE) per user/IP. Redis cluster for HA. Trade-off: Redis latency adds ~1ms per request. Alternative: token bucket with local approximation + periodic Redis sync (leaky bucket). Flipkart deals with this for seller API throttling.
Tell me about a time you improved the performance of a system you were working on.
Tip: Metrics matter: "reduced page load from 4s to 1.2s" beats "made it faster." Walk through: profiling to find bottleneck → hypothesis → change → measurement. Flipkart values engineers who instrument first and optimise based on data, not intuition.
What is eventual consistency in distributed systems? When is it acceptable?
Tip: Eventual consistency: replicas will converge to the same state given no new updates — but reads may see stale data. Acceptable: shopping cart (item count can be stale), product reviews, recommendation scores. NOT acceptable: inventory (overselling is costly), order status, payment confirmation.
What challenges do you anticipate when working on Flipkart's Big Billion Days scale?
Tip: Expected: 10× normal traffic within minutes. Challenges: autoscaling latency, cache stampede on flash sales, database hotspots on popular SKUs, payment gateway saturation. Preparation: load testing, feature flags for graceful degradation, pre-warming caches, circuit breakers on downstream services.
Write a function to detect if a linked list is a palindrome.
Tip: Find middle (slow/fast pointers), reverse second half, compare with first half, restore the list. O(n) time, O(1) space. Alternative: copy to array and use two pointers — O(n) space. Flipkart interviewers expect the in-place O(1) approach.
Flipkart Internet Pvt. Ltd. interviews follow a 4-round process. Here is what to expect and how to prepare for each stage.
Upload your resume and get questions scored across technical depth, communication, structure, confidence, and relevance — the same criteria Flipkart panels use.