Startup · Bengaluru
Meesho interviews test backend scalability, data engineering, and first-principles thinking for their social commerce platform.
Process: Online Coding → Technical × 2 → System Design → HR
What is Meesho and what makes their engineering challenges unique?
Tip: Meesho: social commerce platform connecting resellers and suppliers in tier-2/3 India. Unique challenges: extremely low-bandwidth users (2G/3G), vernacular language support (12+ Indian languages), 10M+ SKUs from unorganised suppliers (noisy catalog data), and cash-on-delivery (COD) logistics optimisation.
How would you design Meesho's product catalog system for 50M+ SKUs from unstructured supplier data?
Tip: Ingestion: supplier uploads images + text → ML pipeline (product categorisation, attribute extraction, duplicate detection). Storage: Elasticsearch for search, S3 for images, Postgres for structured metadata. De-duplication: product matching model using embeddings (same product from multiple suppliers). Quality scoring to rank catalog completeness.
Write code to find the maximum profit from buying and selling a stock with at most K transactions.
Tip: DP: dp[k][i] = max profit using at most k transactions up to day i. Transition: dp[k][i] = max(dp[k][i-1], max over j < i of: price[i] - price[j] + dp[k-1][j]). Optimise: track max(dp[k-1][j] - price[j]) as a running variable. O(kn) time, O(kn) or O(n) space with rolling array.
How does Meesho handle COD (Cash on Delivery) logistics at scale?
Tip: COD is 60–70% of Meesho's orders — unique challenges: cash collection at delivery, COD return fraud (order placed never intended to collect), remittance delays (3PL holds cash before transferring). Engineering: COD risk scoring model, automated remittance reconciliation, fraud signal (repeat non-delivery addresses).
Explain how you would implement multilingual search for Meesho's 12+ language catalog.
Tip: Approach: (1) Transliteration: Hinglish → Hindi (user types "kurti" → also searches "कुर्ती"). (2) Translation layer: query → English canonical form → Elasticsearch. (3) Language-specific analyzers in Elasticsearch (devanagari tokeniser). (4) Pre-translate popular queries. Meesho has a dedicated NLP team for this.
What is an inverted index and how does Elasticsearch use it for Meesho's search?
Tip: Inverted index: maps each unique term to the list of document IDs containing it. Forward index: document → terms. Inverted: term → documents. ES builds this during indexing. For "kurti blue cotton": each token indexed separately, intersected at query time. Allows searching 50M SKUs in milliseconds.
Tell me about a time you had to make a system work for low-bandwidth or low-end device users.
Tip: Meesho's users are on 2G and ₹5,000 Android phones. Real engineering empathy here. Examples: image compression, lazy loading, reducing bundle size, building progressive web apps, offline-first architecture, prefetching on WiFi. Show you understand that 'works on my machine' is not good enough.
How would you build Meesho's reseller commission tracking and payout system?
Tip: Event: order delivered → trigger commission event → write to a ledger table (credit reseller_id, amount, order_id). Payout: weekly batch: aggregate ledger → create payout record → initiate NEFT/UPI transfer → mark settled. Idempotency: payout_id as unique key prevents double payout. Audit trail: immutable ledger rows.
What is the difference between horizontal and vertical scaling? When does each apply to Meesho's use case?
Tip: Vertical: add more CPU/RAM to existing server (scale up). Horizontal: add more servers (scale out). Meesho: horizontal scaling for stateless API servers (just add more pods), vertical for single-node DB until read replicas/sharding becomes necessary. Horizontal is preferred for Meesho's bursty traffic (sale events).
How do you ensure data consistency in Meesho's order management across multiple services?
Tip: Saga pattern: each service publishes events and listens to events from others. Choreography-based: OrderService publishes ORDER_CREATED → InventoryService listens, reserves stock, publishes INVENTORY_RESERVED → PaymentService listens. Compensating transactions for rollback. Outbox pattern ensures DB write + event publish are atomic.
Fashnear Technologies (Meesho) 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 Meesho panels use.