Swiggy Senior Data Scientist / ML Interview Guide
Swiggy's senior DS / ML interviews cluster around three problem domains: logistics ML (delivery time prediction, route optimisation), recommendation (homepage, restaurant ranking), and pricing (surge, promotions). The loop is shorter than Flipkart and more case-heavy than Razorpay. This guide unpacks the structure and the MLE Path mapping.
Swiggy's senior DS / ML interview is calibrated around three distinct problem domains: logistics ML (delivery time estimation, dispatch optimisation, courier supply forecasting), recommendation (restaurant ranking, homepage personalisation), and pricing (surge pricing, promotion targeting). Candidates are often pre-routed to a specific team, and the loop adapts to the team's domain.
The loop is shorter than Flipkart's (4-5 rounds vs 5-6) and more case-heavy than Razorpay's (live problem-solving in 2 of the 4-5 rounds).
The loop structure (Senior DS / ML, L5)
Standard loop: recruiter call → ML + statistics fundamentals → ML system design + case → A/B testing + experimentation case → behavioural + hiring manager.
Swiggy distinguishes itself with the A/B testing round. Most ML candidates underestimate this round; it's specifically calibrated to test experimentation judgment which is core to Swiggy's product culture.
Round-by-round breakdown
Round 1 — Recruiter call. Standard.
Round 2 — ML + statistics fundamentals. 60 minutes. Bias-variance, regularisation, evaluation, A/B testing basics, common statistical tests. Swiggy probes statistical rigor more heavily than other unicorns — expect questions on power analysis, sample size calculation, confidence intervals.
Round 3 — ML system design + case. 75 minutes. Two parts. First 30 min: design a delivery time estimation model end-to-end. Second 45 min: a case study around the design — "your delivery ETA model is off by 4 minutes on average. Walk through your diagnosis." The case round specifically tests production diagnostic judgment.
Round 4 — A/B testing + experimentation case. 60 minutes. "We launched a new ranking model. It shows +2% CTR but -1.5% order completion rate. Do we ship?" Followed by deep probing on: how do you decompose the metric movement, when is the result statistically significant, what guard-rail metrics matter, what causes a metric change without a treatment effect, what are SUTVA violations, how do you handle ranking experiments with marketplace effects.
Round 5 — Behavioural + hiring manager. 45 minutes. Team fit, project ownership, leadership stories.
What Swiggy weights distinctively
Three things:
1. Statistical rigor. Swiggy interviewers grade more heavily on power analysis, confidence interval interpretation, and multiple-testing corrections than most Indian unicorn loops. The product culture is experimentation-first. 2. Marketplace dynamics. Swiggy is a two-sided marketplace (customers + restaurants + delivery partners). Senior candidates are expected to articulate how ML decisions on one side affect the others. Recommending restaurants affects supply-side ordering; surge pricing affects partner availability. 3. Logistics ML specifically. Time-series forecasting, real-time dispatch, supply-demand prediction. The MLE Path Tier 6 (Sequence & Specialised) maps directly to this.
Top 10 questions Swiggy senior DS / ML candidates face
1. "Design Swiggy's delivery ETA model. What features, what target, what loss function?" 2. "Your ETA model is off by 4 minutes on average. Three hypotheses in order." 3. "A new ranking model shows +2% CTR but -1.5% order completion. Do we ship?" 4. "How do you A/B test a ranking model when one user can see different rankings within a session?" 5. "Design surge pricing for the dinner rush. What signals drive the multiplier, what guards against gaming?" 6. "Walk me through the most impactful experiment you've designed and run." 7. "How do you handle the marketplace-spillover effect when running an experiment on a recommendation model?" 8. "Sample size calculation for an A/B test where the expected lift is 1%. Walk through it." 9. "Your model predicts 8 minutes; the actual is 16 minutes. The customer cancels. What do you fix — model, threshold, or product surface?" 10. "Eleven types of leakage. Which apply to delivery time estimation specifically?"
The prep path through MSL
For a Swiggy senior DS / ML loop:
For practice: CausalInferenceTab (experiment design + SUTVA), MLCodingTab, Mock Interview with a Swiggy JD pasted.
Common failure modes
Compensation
Swiggy Senior DS / ML in 2026 ranges roughly ₹40 lakh – ₹70 lakh for 5-7 YOE. Staff scales to ₹85 lakh+. AmbitionBox and Glassdoor have public ranges.
Final note
Swiggy's loop tests for statistical rigor + marketplace judgment + logistics ML. The MSL Path Tiers 1, 5, 6, plus a cross-link to PAL, is the prep map.