Machine learning roles in India have expanded beyond pure research into applied ML, MLOps, and AI product engineering. Indian product companies, BFSI AI teams, and a growing number of AI startups are hiring ML engineers and data scientists who can build, evaluate, and deploy models in production. This guide covers the ML interview questions asked at these companies in 2026.
Core concepts tested at every level: Supervised vs unsupervised vs reinforcement learning. Bias-variance tradeoff: high bias = underfitting, high variance = overfitting. Cross-validation: k-fold to estimate generalisation error. Regularisation: L1 (Lasso: sparse weights, feature selection) vs L2 (Ridge: shrinks all weights). Gradient descent: batch, stochastic, mini-batch. Learning rate schedules. Overfitting detection and remedies: regularisation, dropout (neural nets), more data, early stopping. Feature scaling: StandardScaler for gradient-based models, MinMaxScaler for distance-based models (KNN, SVM): tree-based models do not need scaling.
Algorithms tested at Indian ML interviews: Linear and logistic regression (assumptions, interpretability, when to use). Decision trees (information gain, Gini impurity, pruning). Random forest (ensemble of trees, out-of-bag error, feature importance). Gradient boosting: XGBoost, LightGBM, CatBoost: the go-to for structured/tabular data at Indian product companies. SVM: kernel trick, support vectors, C and gamma parameters. KNN: lazy learner, curse of dimensionality. Neural networks: architecture (layers, activations), backpropagation, common architectures (CNN for images, LSTM/Transformer for sequences). For recommendation systems (Flipkart, Zomato, Swiggy use cases): collaborative filtering, matrix factorisation, content-based filtering.
Model evaluation metrics are heavily tested: interviewers expect you to know which metric to use when and why: Accuracy: misleading for imbalanced datasets. Precision: out of predicted positives, how many are correct? Recall (Sensitivity): out of actual positives, how many did we catch? F1 score: harmonic mean of precision and recall: use when both matter. ROC-AUC: rank-ordering ability, threshold-independent: use to compare models. PR-AUC: better than ROC-AUC for highly imbalanced datasets (fraud detection, rare disease prediction). RMSE, MAE for regression: MAE is more robust to outliers. Business metric alignment: always connect model metrics to business outcomes: reducing churn prediction false negatives reduces revenue loss.
MLOps questions appear at senior ML roles at Indian product companies: Feature stores: centralised feature management (Feast, Tecton): prevents training-serving skew. Training-serving skew: features computed differently during training vs inference: the most common production ML bug. Model versioning: MLflow, Weights and Biases. CI/CD for ML: automated retraining pipelines, data validation (Great Expectations), model performance monitoring (data drift, concept drift). A/B testing ML models: holdout groups, traffic splitting, statistical significance. Model serving: REST API (Flask/FastAPI), model as a microservice, ONNX for cross-framework deployment, TorchServe, TensorFlow Serving. Shadow deployment: running a new model in parallel to compare against production before switching.
Machine learning roles in India saw significant compensation growth through 2025 and into 2026. Entry-level ML engineers at service companies earn 7 to 12 LPA, while data scientists at product startups like Dunzo, Ola, or Udaan start at 15 to 22 LPA. Senior ML engineers at Flipkart, Amazon India, or Myntra's AI teams command 35 to 55 LPA. Specialist roles in generative AI, LLM fine-tuning, and recommendation systems at firms like Sarvam AI, Krutrim, or Google DeepMind India reach 80 to 120 LPA. Professionals with publication records or Kaggle Grandmaster status often negotiate 20 to 30 percent above standard bands.
Flipkart's data science team runs four rounds: an online assessment with SQL and statistics questions, a machine learning fundamentals call, a case study round requiring end-to-end model design for recommendation or pricing, and a hiring manager discussion on business impact. Swiggy and Zomato focus heavily on time-series forecasting and demand prediction problems in their interviews, reflecting their core delivery operations. Ola's AI team often presents candidates with a dataset and asks them to walk through feature engineering decisions live. CRED interviews include probability puzzles and A/B testing design. Most Indian product companies now also include an LLM or prompt engineering component regardless of the primary role focus.
Interviewers at Indian companies frequently present business problems rooted in local market realities. Common case studies include building a credit scoring model for thin-file customers in tier-2 cities, designing a vernacular language NLP pipeline for Hindi or Tamil product reviews, or building a fraud detection system for UPI transactions that processes tens of millions of events daily. Candidates should be familiar with class imbalance techniques like SMOTE given that fraud datasets are highly skewed. Knowledge of Indic language models from AI4Bharat, such as IndicBERT or Dhruva, signals genuine India-market awareness and is increasingly valued at companies serving non-English-speaking users across Bharat.