Expect applied ML design (predicting delivery delays at order time, including features, model choice, retraining cadence, and feedback loops), diagnosis questions (a pricing model that looks good offline but underperforms live), and experiment-design questions where network effects break naive A/B tests. Interviewers reward candidates who reason about distribution shift, train-serve skew, feedback loops, and interference rather than just naming algorithms. Flipkart's data science powers pricing, delivery, and ranking, so grounding answers in a real e-commerce workflow with a working feedback loop stands out. Prepare ML system design, evaluation pitfalls, and experimentation.
Recruitment process
- Recruiter screen and technical pre-screen
- ML fundamentals and coding round
- ML system design and experimentation round
- Hiring-manager and behavioural round, then offer
Interview rounds
- Round 1 (45-60 min): ML fundamentals and coding round
- Round 2 (60 min): ML system design round with feedback loops
- Round 3 (45-60 min): experimentation and model-diagnosis round
- Round 4 (45 min): behavioural and hiring-manager round
End to end and live framework
- Design ML systems with feature, retraining, and feedback loops, diagnose the offline-online gap, and handle interference in experiments at e-commerce scale
Practice questions
- 01Practise
Reverse a linked list in O(1) extra space. Then explain when you'd use this in production.
Standard - 02Practise
Given a large product catalogue, return the top-K best-selling items in a category efficiently as new sales stream in. Which data structure, and why?
IntenseFlipkart technical rounds pair classic DSA with a machine-coding round; problems are usually framed in commerce terms (catalogue, cart, inventory). Expect a heap or count-min discussion and a follow-up on updates. - 03Practise
Design and code an in-memory rate limiter or LRU cache for a product-detail service. Keep it clean, testable, and thread-safe.
IntenseFlipkart's machine-coding round rewards clean, extensible OOP with clear interfaces and unit-testable structure over cramming everything into one function. - 04Practise
Given delivery slots and orders, assign orders to slots to maximise fulfilment without overbooking any slot. Walk through your approach and complexity.
Standard - 05Practise
Write a SQL query to find, for each product category, the top 3 sellers by revenue in the last 30 days, handling ties consistently.
StandardFlipkart data-engineer screens lean heavily on window functions. Expect ROW_NUMBER/RANK, partitioning, and correct tie-handling. - 06Practise
Design a batch and streaming pipeline that ingests clickstream events and produces near-real-time category-level conversion metrics. How do you handle late-arriving events?
IntenseFlipkart probes lambda/kappa trade-offs, watermarking, and idempotent aggregation for e-commerce analytics at scale. - 07Practise
How would you model an orders fact table and its dimensions for an analytics warehouse serving both finance and category teams? Discuss grain and slowly changing dimensions.
StandardFlipkart data-modelling rounds test dimensional modelling: fact-table grain, star schema, and SCD type-2 handling. - 08Practise
Design an ML system to predict delivery delays at order time. What features, model, and feedback loop would you use, and how do you avoid stale predictions?
IntenseFlipkart data-scientist rounds test end-to-end ML system thinking: feature engineering, model choice, retraining cadence, and feedback loops. - 09Practise
A pricing model performs well in offline tests but underperforms live. List the likely reasons and how you'd investigate each.
StandardFlipkart probes the offline-online gap: distribution shift, feedback loops, leakage, and train-serve skew, diagnosed systematically. - 10Practise
How would you design an A/B test for a new ranking model when network effects mean one user's experience affects another's?
IntenseFlipkart data-scientist rounds probe experiment design under interference: cluster randomisation and handling network effects.
