Flipkart vs Amazon India Interview: Two Very Different Tests

You're applying to both Flipkart and Amazon India at the same time. Smart. Efficient. Also, a little chaotic. Because these two companies, which operate in the same country selling some of the same products, have designed interview loops that have almost nothing in common beyond "there will be code."

Flipkart's signature is a 90-minute machine coding round where you build a working object-oriented system from scratch, submit a zip file, and pray. Amazon India's signature is a behavioral track woven through every single round, where the 16 Leadership Principles carry as much weight as the algorithm, and a Bar Raiser from a different team can sink an otherwise-strong loop if your STAR stories don't hold up when someone starts asking follow-up questions.

Same domain, different tests.

Flipkart vs Amazon India: Rounds at a Glance

Stage	Flipkart	Amazon India
Online Assessment	90 min, 3 DSA problems (HackerRank)	90 min: 2 technical + 20 min systems design + 8 min Work Style Survey (source)
Technical Round 1	PSDS-1: DSA, 60 min, Google Doc	55 min: technical + 2-3 behavioral questions
Technical Round 2	PSDS-2: DSA (if PSDS-1 passes), 60 min	55 min: technical + 2-3 behavioral questions
Signature Round	Machine Coding: 90 min code + 30 min review (source)	Bar Raiser: LP-heavy, may have no coding
Design Round	System Design (SDE-2+): HLD + LLD	System Design (SDE-2 loop): HLD
Final	Hiring Manager + HR	Hiring Manager

End to end, Flipkart takes three to five weeks. Amazon's loop typically wraps in two to three weeks after the OA.

Inside the Flipkart Loop

The PSDS Rounds Are Warmer Than You Expect

Flipkart's Problem Solving and Data Structures rounds happen in a shared Google Doc. No compiler, no IDE, no autocomplete. Just you, the interviewer, and a blank text editor with no syntax highlighting.

The interviewer is evaluating your thinking process, not your ability to pass a compiler. Pseudocode is fine if the reasoning lands. If your first approach is naive, expect harder follow-ups. Don't coast.

Common patterns: sliding window, two pointers, prefix sums, trees, heaps, DP. Difficulty lands at LeetCode medium-hard. PSDS-1 typically has one easier and one harder problem in 60 minutes. PSDS-2 is gated on passing PSDS-1 and usually dials up the difficulty. A representative PSDS-2 problem: "design an LRU cache with O(1) get and put." It sits at the boundary between algorithm and LLD, exactly where Flipkart likes to live. For the playbook see the DSA patterns cheat sheet.

The Flipkart Machine Coding Round Is the Real Filter

This is the round that defines Flipkart's loop. You get one object-oriented design problem and 90 minutes to write working, extensible code, followed by a 30-minute review with the interviewer (GeeksforGeeks: Flipkart Machine Coding Round Experience confirms this 90+30 split).

The rules:

Any language is fine (most candidates use Java or Python)
No external databases. Everything in-memory: HashMap, ArrayList, HashSet
You submit a zip file via a Google Form when time is up
The code must run. It must also be readable and defensible

Recent problems from the machine coding round include:

Coding Blox: an online competitive programming platform where users sign up, create contests with difficulty tiers, and participate in contests run by others
Wallet System: load money, send to users, check balance, transaction history with sorting and filtering
Doctor Appointment Booking: doctors declare availability, patients search by specialty and book
Distributed Task Scheduler: workers pick up tasks, dynamic load balancing, task retry on failure

Every problem is a small production system, not a trick puzzle. You're building something that could ship.

Flipkart prioritizes running code over pristine architecture. If you have to choose between a working solution with messy design patterns and a beautiful but broken design, submit the working one. The 30-minute review is where you defend your choices, so at least understand why you made each decision.

Four design patterns recur across these problems. Map each one to the problem it solves:

Strategy maps to Coding Blox. Different contest types (rated, unrated, ICPC-style, ladder) score submissions differently. One ScoringStrategy interface, one concrete class per rule set.
State maps to Doctor Appointment Booking. A slot moves from Available → Booked → Completed (or Cancelled). The legal next actions depend on the current state, so model state, don't write nested ifs.
Factory maps to the Wallet System. A TransactionFactory.create(type) lets you spawn Deposit, Transfer, Refund, or Reversal without the caller knowing the constructor signature of each.
Observer maps to the Distributed Task Scheduler. Workers subscribe to a task queue; when a new task lands or a retry is scheduled, every idle worker hears about it without polling.

You don't need the GoF book. You need a one-line trigger from each problem signal to a pattern.

A meme showing a programmer confidently submitting code labeled "it works!! don't touch!!!" with chaotic internals

Every Flipkart machine coding submission after minute 80.

For Flipkart's full process, see the Flipkart software engineer interview guide.

System Design Wants You to Drive the Scope

For mid-level roles you get both LLD and HLD in one round. Drive the scope discussion yourself. Ask about user scale, geographic reach, and consistency requirements before touching any diagram. LLD first: class schema, database tables with normalization, API signatures. HLD follows: service decomposition, Kafka queues, caching layers. Be ready to defend every schema decision. If you put something in a table, know why it's there and not somewhere else.

Inside the Amazon India Loop

The OA Has Three Distinct Parts (One of Them Will Surprise You)

The Amazon SDE-II online assessment runs about 90 minutes and has three sections: two technical coding questions, 20 minutes of systems design scenarios, and an 8-minute multiple-choice Work Style Survey (amazon.jobs SDE-II prep).

The Work Style Survey trips people up because it looks trivial. It isn't. Amazon uses it to filter for Leadership Principle alignment before you ever talk to a human. Scenarios include ambiguous email chains, escalating conflicts, and tradeoff decisions under resource constraints. You pick the most and least likely action you'd take. There's no "right answer" listed. There is, however, a right answer.

Every Technical Round Also Has LP Questions. Every. Single. One.

Amazon's own SDE-II guidance says each interviewer "will typically ask two or three behavioral-based questions about successes or challenges" (amazon.jobs). Technical and behavioral questions happen in the same 55-minute round, usually 25-30 minutes on DSA and the rest on behavioral. The behavioral portion isn't an afterthought. It's half the round.

At Amazon, the LP signal can override a strong technical signal. Both dimensions are formally scored and both feed into the debrief. If your STAR answers are weak, your graph traversal won't save you. Practice the framing in the behavioral interview cheat sheet and the technical interview communication guide.

DSA at Amazon India trends toward LeetCode medium, with occasional hard problems for SDE-2. Common patterns: graphs, hash maps, sliding window, trees. Problems usually have an implicit systems context. See the Amazon software engineer interview guide for the full breakdown of rounds, levels, and what each interviewer is scoring.

The Bar Raiser Round Is the Final Boss

The Bar Raiser is a trained interviewer from a different team. Per Amazon's official description, there are 3,600+ Bar Raisers across the company and their job is to ensure every new hire is "better than 50% of their would-be peers in similar roles." They sit on the debrief as an objective third party with the authority to push back on the hiring manager. Think of them as the appeals court of the hiring loop, not a final round.

The Bar Raiser round is often entirely behavioral. No coding. Two LPs in extreme depth, with follow-ups designed to stress-test your stories.

Here's what the depth gap looks like. Same prompt: "Tell me about a time you took ownership of something outside your scope."

Shallow:

"Our payment service was timing out. I raised it with my manager, we migrated, timeouts went down."

Deep:

"During our Big Billion Days dry run, our payment gateway hit 30% timeout rate at 8K RPS. I owned the call to migrate to the secondary gateway: one-page comparison of three options, design review with the platform team, a rollback playbook with a 90-second revert SLA, cutover at 2am Sunday. Timeouts dropped to 0.4%. Two follow-on changes: a synthetic 5x-load canary in pre-prod, and an SOP for gateway swap any on-call SDE can run."

The Bar Raiser will chain follow-ups on the deep version:

"Why was this your call? Where was your skip?"
"Walk me through the exact rollback. Which branch, which command, which dashboard?"
"What was the result six months out, not the day of?"

A rehearsed but shallow STAR story dies on the second follow-up. A real one keeps going. The trick isn't more stories. It's three or four stories you actually lived, where you remember the proper nouns. Different LPs can draw from the same story as long as the emphasis shifts. The three that come up most for SDEs: Ownership, Deliver Results, and Invent and Simplify.

A meme labeled "conflictResolved" showing a confident programmer after solving a conflict

Bar Raiser: "Tell me about a time you had a conflict with a teammate." Your STAR story, if you prepped:

DSA: How Do the Difficulties Compare?

Both companies aim at LeetCode medium-hard, but they ask the question differently. Flipkart hands you the textbook prompt. Amazon hides the textbook prompt inside a story.

Flipkart's PSDS rounds favor clean algorithmic problems. Representative ask: "given a stream of integers, support addNum and findMedian in O(log n) and O(1)." The answer is a two-heap structure. The interviewer often nudges you if you're stuck. Adversarial in difficulty, not in spirit.

Amazon's technical rounds layer systems context onto the same algorithm. Representative ask: "we process click events from our retail site. Return the top K most-viewed SKUs in the last hour at any time, under millisecond latency." Underneath, that's a sliding window plus a heap, but you have to extract it from the scenario first. The skill being tested is recognizing the pattern through the story.

For both companies, these patterns matter most: arrays and strings, trees and graphs, dynamic programming, heaps, and hash maps. For Amazon, also practice problems with a stream or event-driven framing. For Flipkart, practice problems that live at the LLD/algorithm boundary (LRU cache, task scheduler internals, rate limiting).

The DSA guide for backend engineers covers the patterns most relevant to both loops.

How to Prepare for Each

Preparing for Flipkart

Spend at least four weeks on machine coding. Pick a problem (appointment booking, parking lot, library management), open a blank editor, build it end to end with a 90-minute timer, then review what you'd change. The first time I tried this I burned 35 minutes on the class hierarchy and wrote zero working methods. Lesson: start with the dumbest end-to-end skeleton and refactor as you go. You can't defend a design that doesn't run.

For DSA, 30-40 LeetCode mediums across the core patterns is enough. Brush up on database normalization. The SDE-2 system design round probes 1NF/2NF/3NF when you defend your schema.

Preparing for Amazon India

Write your LP stories before you touch a LeetCode problem. This is the part candidates skip, and it's the part that kills them.

For each of the 16 LPs, write a STAR story from your work history. Five to six strong ones that flex across multiple principles is enough. A production bug fix can serve Ownership, Deliver Results, or Bias for Action depending on what you highlight.

Practice answering LP questions out loud. Bar Raiser-style follow-ups can run for several minutes per story, and Amazon's own Bar Raiser tips emphasize ending with results and showing what you learned from a failure (aboutamazon.com). If you've only rehearsed the written version, you won't survive the real version. A tool like SpaceComplexity can help you run realistic mock interviews where you get rubric-based feedback on your spoken communication, not just whether you got the algorithm right.

For DSA, 50 LeetCode mediums across graphs, trees, arrays, and hash maps is the baseline.

A Six-Week Timeline for Both

Weeks 1-2: DSA mediums (shared prep)
Week 3: LP stories written and drilled out loud
Week 4: Machine coding with timer, two to three full problems
Week 5: System design concepts, mock LP interviews
Week 6: Full mock interviews for both formats

Which Should You Target First?

Target Flipkart first if you have strong OOP instincts but no behavioral story bank yet. A great machine coding round can carry you past a mediocre DSA round, and there's no LP equivalent waiting on the debrief.

Target Amazon first if you have strong stories and solid DSA but haven't practiced LLD-heavy machine coding. Amazon's loop has more room to recover from one weaker round because you're scored across many dimensions.

Run them in parallel when you can. Recruiting timelines in India often overlap.

Key Takeaways

Flipkart: 90-min machine coding plus 30-min review is the gate. Working > pristine
Amazon India: LPs run through every round, plus a Bar Raiser from another team on the debrief
DSA is LeetCode medium-hard on both, but Amazon hides the pattern inside a scenario
Strategy/State/Factory/Observer cover most Flipkart problems. Map each pattern to a signal
Write LP stories before LeetCode. Shallow rehearsed stories break on the second follow-up