Inmobi

Round 0 Type: Written Round This was an Online Assessment (OA). It consisted of 1 DSA question and 25 MCQs. The DSA question was graph-based. You were given integers n and t. A graph was defined such that there is an edge between nodes i and j if gcd(i, j) > t. The task was to find: The number of connected components The size of the largest component The remaining 25 questions were MCQs covering: Probability and Statistics Central Limit Theorem (CLT) Tree-based ML models LoRA Transformers Other ML-related concepts -- Round 1 Type: Technical Interview This round was an in-depth discussion about my ongoing internship at Amazon as an Applied Scientist Intern. The entire discussion focused on the work I was doing there, including the problems I was solving, methodologies used, and my contributions. -- Round 2 Type: Technical Interview This round focused on breadth in Machine Learning concepts. I was asked about: Tree-based ML models Core ML fundamentals The interviewer then went into deep technical details about the Transformer architecture. The discussion around Transformers concluded the round. -- Round 3 Type: Hiring Manager (HM) Round This round started with a scenario-based question. I was asked: If all weights in a neural network are initialized with the same value, what would happen? How would the weights update during training? This was essentially a backpropagation and symmetry-breaking analysis. Next question: Given a biased coin where P(Tail) = 0.1, how would you make this coin unbiased? After that, there were more ML concept-based questions: Attention mechanism Batch Normalization (what it is and why we use it) Precision and Recall Stochastic Gradient Descent vs Mini-batch Gradient Descent (comparison) Tree-based ML models Variance aspects in Bagging vs Boosting Then there was a scenario-based question: You are given a reference image Ri. You need to recommend similar images from a catalog to the user. The discussion revolved around designing an end-to-end system, including feature extraction, embeddings, similarity measures, and retrieval strategy. Finally, I was asked: If you receive an offer from Amazon, why would you choose inMobi? -- Round 4 Type: HR Round This was a general discussion-based HR round. -- Sorry for the earlier poor English. I hope this experience write-up helps.