Interview Prep · ML Systems Lab

InMobi / Glance ML Engineer Interview Guide

InMobi (and its consumer brand Glance) runs ML at advertising scale. The interview emphasises CTR prediction, ad targeting, real-time bidding, and personalisation. Closest comparison is to Google Ads ML interviews but with Indian-market context. This guide unpacks the structure and prep.

InMobi's ML organisation runs at advertising scale — billions of ad impressions, real-time auction decisions in single-digit milliseconds, complex multi-objective optimisation balancing advertiser ROI, publisher revenue, and user experience. The interview reflects this.

The loop structure (Senior ML Engineer)

Standard loop: recruiter screen → ML fundamentals + coding → ad ML system design → real-time bidding case → hiring manager.

Round-by-round breakdown

Round 1 — Recruiter screen. Standard.

Round 2 — ML fundamentals + coding. 60 minutes. Conceptual probing with emphasis on calibration (ad CTR predictions feed bid prices, calibration matters more than ranking quality), AUC vs PR-AUC at imbalanced data (CTR is < 5%, often < 1%), and online learning fundamentals (advertising ML retrains frequently).

Round 3 — Ad ML system design. 60 minutes. Common prompts: design a CTR prediction system for display ads, design real-time bidding for ad auctions, design personalised ad targeting. The design has to address feature engineering (user features, ad features, context features, cross features), serving (sub-10ms latency for bidding), and feedback loops (training on logged ad data has bias).

Round 4 — Real-time bidding case. 60 minutes. Real scenario: "your CTR model's calibration shifted overnight. Bids are now 30% too high or 30% too low across the board. Diagnose and fix." Tests the discipline of production calibration monitoring.

Round 5 — Hiring manager. 45 minutes.

What InMobi weights distinctively

1. Calibration depth. Ad bidding is calibration-bound. InMobi probes calibration much more deeply than most Indian unicorn loops. 2. Auction theory awareness. Ad ML candidates are expected to understand auction mechanics (first-price vs second-price, bid shading, reserve prices). 3. Online learning patterns. Models retrain on streaming data. Senior candidates are expected to articulate online learning trade-offs.

Top 10 questions InMobi candidates face

1. "Design CTR prediction for display ads. CTR base rate is 0.3%." 2. "Your bid model's calibration shifted. 30% too high or too low. Diagnose." 3. "How does first-price vs second-price auction change your bidding strategy?" 4. "Design real-time targeting. What features can you compute in 10ms?" 5. "Walk through your most impactful ad ML project." 6. "Click-baiting feedback loop: your model learns to recommend click-baity content. How do you detect and fix?" 7. "Online learning vs batch retraining for CTR. Trade-offs." 8. "Eleven types of leakage. Which apply to ad CTR prediction specifically?" 9. "How do you A/B test a bidding model when the bid changes the population of users you serve?" 10. "Cold start for a new advertiser. What signals do you use?"

The prep path through MSL

For InMobi / Glance:

  • Tier 3 (Classical Algorithms) — Post 76 (Calibration) and Post 129 (Class Imbalance) are central.
  • Tier 5 (Evaluation & Diagnostics) — Post 3 (AUC critique) and Post 130 (Leakage Taxonomy).
  • Tier 7 (Production Engineering) — all 5 posts.
  • Tier 8 (Monitoring & MLOps) — Post 40 (Calibration Loss in Production) is exactly the round 4 prompt.
  • Tier 9 (System Design) — Post 24 (6-Step Framework), Post 80 (Semantic Search for targeting).
  • Compensation

    InMobi / Glance ML Engineer in 2026 ranges roughly ₹40 lakh – ₹75 lakh for 5-7 YOE. Staff reaches ₹95 lakh+. InMobi pays competitively for the role family.

    Continue interactively
    Read this post inside ML Systems Lab — with Simplify toggle, interview Q&As, inline glossary, and the MLE Path forward pointer.
    Open in MSL →