Learn from Flipkart SDE-1 interview experiences for campus and off-campus drives. These composites pull patterns that keep showing up - timed 3-problem OAs, DSA-heavy technicals with dry runs, and DBMS/project-aware closing rounds.
Candidate profile: Composite of Medium Sep 2025 campus Intern+FTE and GFG off-campus 2024 SDE-1
Round 1 - Online Assessment (90 min)
Campus Medium: 3 DSA offline; 22 shortlisted after OA on that campus
Off-campus GFG Apr 2024: SHL, 3 medium–high problems; one-time link (cannot reopen) - themes included maximize simultaneous API runtime; word-break style sentence generation
Tip: Start SHL only with a quiet 90 minutes - no pause/reschedule in that report
Round 2 - Technical (~60 min)
Medium account: deep ML price-recommendation project (scraping, stack, GitHub demo) then live Rotten Oranges / grid BFS with platform test cases - accepted
12 students to Round 2 in that write-up; candidate called within minutes of Round 1
Off-campus Tech 1: two medium–high coding; DP + hash maps; second problem may clear on strong pseudocode + approach
Candidate profile: Composite of public SDE-2 / lateral reports (InterviewExperiences.in Feb 2026; LinkedIn flight-booking experience; Medium SDE-2 notes). Campus SDE-1 may see a shortened variant - confirm your drive mail.
Round 1 - Machine coding (90–120 min)
Problem delivered via Google Doc or portal; sometimes a 30-minute explainer session first
In-memory console app - no real DB or HTTP; driver/main with test cases required
Reported domains: quick-commerce delivery (customers, partners, orders); flight booking (search, sort by cost/time, routes); doctor appointments (slots, speciality search, double-booking rules); distributed task scheduler with load balancing
Grading emphasis (candidate-reported): demoable correctness, readability, entity modeling, extensibility, exception handling - bonus features only after P0 flows run
Submit as zip via form; AI assistants often prohibited; internet for syntax only
Tip: Ship add/remove/search happy path in 45 minutes, then one discount/rule extension + one test
Round 2 - Machine coding evaluation (30–45 min)
Walk through class diagram and design patterns used (Strategy, Factory, Observer reported)
“What if two tasks/orders arrive concurrently?” - locking or queue discussion
Production follow-ups: persistence, API layer, metrics - you are not expected to code them, only articulate
Result: Advanced when code ran and candidate owned trade-offs honestly
Later rounds (role-dependent)
DSA or system design for senior loops; HM behavioural on ownership and incidents
Final result: Offer when evaluation + later bars clear
Takeaways
What decided it: Treat machine coding as two interviews - build and defend. Simple running design beats elaborate UML that does not compile.
Common fail: Shallow e-commerce design or weak OA edge cases
How to practise: Catalog/cart/inventory design sketches; timed OA with edge cases
Sample answers for high-frequency questions
Use these as frames, not scripts - interviewers probe follow-ups.
Technical
Q. Longest subsequence where max − min = 1? A. Build a frequency map. For each value x, best length using only x and x+1 is freq[x] + freq[x+1]. Track the maximum. Time O(n), space O(n).
Q. Pascal triangle element at row n, index k? A. Compute C(n, k) with the multiplicative formula - iteratively multiply (n−i)/(i+1) to avoid full factorial overflow. Clarify 0-based vs 1-based indexing with the interviewer.
Q. Rotten Oranges (multi-source BFS)? A. Enqueue all initially rotten cells; each minute spoil adjacent fresh oranges. Track remaining fresh count; if fresh remains when the queue empties, return -1, else minutes elapsed. Time O(R×C).
Q. SQL vs NoSQL / MongoDB? A. SQL: relational integrity, joins, transactions - orders and payments. NoSQL documents: flexible catalogues, high write throughput, horizontal scale when join-heavy schemas hurt. MongoDB stores BSON documents in collections - mention indexes and when denormalization helps reads.
Q. LRU cache - preferred structure? A. Hash map key → node plus doubly linked list for recency. get/put O(1). Array or singly linked list fails cheap unlink of arbitrary nodes.
Q. Machine coding: how do you prioritize in 90 minutes? A. List P0 user stories from the prompt; implement entities + repository in-memory; driver that runs sample tests; only then Strategy/Factory for extensibility. State assumptions aloud for the reviewer.
Q. Word Ladder (BFS)? A. Build graph edges between words one letter apart; BFS from beginWord to endWord; track path/predecessor for reconstruction. Discuss wordLength and alphabet size for complexity.
HR
Q. Why Flipkart / how will you help? A. Name a product surface (search, pricing, logistics), map one skill you already have, admit one gap you will learn, and stress team leverage over lone-hero narratives.
Q. Challenges in a project? A. STAR: specific failure (data quality, latency, scraping blocks), action you owned, result, and process change afterward.
These links are research inputs we used to map round structure and recurring questions - not copied page mirrors. We reorganized reports into process tables, outcome notes, and prep takeaways for Flipkart.
Ready to prepare? Drill timed 3-problem sets with one hard grid BFS, own one demoable project, and revise SQL vs NoSQL until you can debate with examples.
Pro tip: Shortlist counts in campus blogs are local snapshots - use them for pacing intuition, not as predicted cutoffs for your college.
