InMobi Machine Learning Engineer Campus Guide: GCD Graph OA, ML Depth and Transformers
Prepare for InMobi machine learning engineer interviews with reported assessment patterns, interview questions, and preparation advice.
On this page
Last reviewed November 16, 2026.
For students applying for InMobi's ML engineer role on campus.
Built by the Codinzhub team from an InMobi ML engineer campus experience (selected) and InMobi's 2026 campus process. One detailed report exists for this role, so use this as a map, not a promise.
The 30-second version. The ML engineer OA is 1 DSA question and 25 MCQs on probability, statistics, tree models, LoRA and transformers. Then come three technical rounds (your internship in depth; ML breadth and transformers; a hiring manager round with puzzles and an ML system design) and HR. The selected candidate said the interviews went very deep on fundamentals.
Other InMobi roles: SDE 1 · SDE intern and 6-month intern · Machine learning engineer
Round-by-round breakdown
Stage 1: Online assessment
At a glance
- Questions per round: 1 DSA and 25 MCQs
- What was asked: a graph built from GCDs; probability, the central limit theorem, tree models, LoRA, transformers
- Most common: ML theory MCQs
The DSA question. Nodes 1 to n, with an edge between i and j when gcd(i, j) > t. Return the number of connected components and the size of the largest. Approach: for each d from t + 1 to n, join all multiples of d with union-find (multiples of the same d share gcd at least d); numbers with no such divisor stay alone.
Stage 2: Technical round 1, your internship
What to expect. The whole hour was on the candidate's ongoing applied scientist internship: the problems, methods and their own contribution.
Stage 3: Technical round 2, ML breadth
What to expect. Tree-based models, core ML fundamentals, and a long, detailed discussion of the Transformer architecture.
Stage 4: Hiring manager round
Questions the candidate was asked:
- If every weight in a neural network starts at the same value, what happens during training? (Symmetry is never broken; all neurons learn the same thing.)
- A coin lands tails with probability 0.1. How do you get a fair result from it? (Toss twice; HT and TH are equally likely; repeat otherwise.)
- Attention; batch normalisation and why it helps; precision and recall; SGD versus mini-batch gradient descent; variance in bagging versus boosting.
- Design: given a reference image, recommend similar images from a catalogue: feature extraction, embeddings, similarity, retrieval at scale.
- If you had an offer from a big tech company, why would you choose InMobi?
Stage 5: HR
What to expect. A general discussion.
The Codinzhub Next-Round Playbook
- Revise transformers until you can explain attention, positional encoding and LoRA without notes.
- Practise classic ML puzzles (equal initialisation, fair coin from a biased coin).
- Prepare an image or product similarity system design end to end.
About this guide
This guide is researched and written by the Codinzhub team from questions and experiences students shared on Codinzhub and in public interview write-ups, rewritten in our own words. Practice links open the full question on Codinzhub. Interviewed for this role? Share your experience on Codinzhub and we will update the guide.
Keep track of what you’ve learned.