Machine Learning interview questions

Showing 181–210 of 572 results

• Design a Low-Latency Store RecommenderHard

  This question assesses the ability to architect a multi-stage recommendation pipeline that balances low-latency retrieval with high-accuracy ranking in a geo-aware, high-concurrency food delivery context.

  DoorDashData ScientistMachine LearningOnsite
• Design and evaluate an uplift modelHard

  This question evaluates competency in causal inference and uplift modeling, experimental design for randomized treatment allocation, offline evaluation of uplift models, operational decision rules under budget constraints, fairness guardrails, and handling interference and deployment of sequential tests in a data-science context.

  DoorDashData ScientistMachine LearningOnsite
• Design enterprise file recommendations under ACLsMedium

  This problem tests the ability to design a multi-tenant recommendation system that enforces complex security constraints while maintaining low latency.

  DropboxData ScientistMachine LearningTechnical Screen
• Lease Renewal Prediction and Non-Renewal Reason Categorization Take-HomeMedium

  A machine learning take-home on predicting lease renewals from tenant conversations and SQLite tables of residents, maintenance work orders and renewal offers. It covers preprocessing, a metric compared with a random baseline, a test-set prediction file, categorizing non-renewal reasons with gte-base embeddings, and presenting the results.

  EliseAIData ScientistMachine LearningTechnical Screen
• Build a model to predict wine qualityMedium

  This question evaluates predictive modeling and data-science competencies—exploratory data analysis, feature assessment, problem framing (regression vs classification vs ordinal), model selection and validation, evaluation metric choice, and post-model feature-importance interpretation—in the Machine Learning domain, combining conceptual understanding with practical application. It is commonly asked in technical interviews for Data Scientist roles because it tests end-to-end modeling judgment, reasoning about data distributions and variable usefulness, selection of validation and evaluation strategies, and awareness of pitfalls such as collinearity and data leakage.

  EvenUpData ScientistMachine LearningTechnical Screen
• Debug a PyTorch Contrastive LossHard

  Repair a PyTorch contrastive loss whose matrix multiplication contracts incompatible dimensions instead of producing pairwise scores. Validate aligned rank-two tensors and temperature, create the correct batch-by-batch logits, preserve gradients, and handle low-precision numerical stability.

  ExaMachine Learning EngineerMachine LearningTechnical Screen
• Validate and monitor ranking model end-to-endHard

  This question evaluates a data scientist's mastery of learning-to-rank concepts, offline evaluation and metric design, position-bias correction, diagnostic analysis, deployment safeguards, and alignment of surrogate objectives with business KPIs for hotel search ranking.

  ExpediaData ScientistMachine LearningOnsite
• Explain PCA and L2 Normalization in Machine LearningMedium

  This interview question evaluates core ML concepts, assumptions, math intuition, training/evaluation trade-offs, and practical failure modes in a realistic interview setting. A strong answer for Explain PCA and L2 Normalization in Machine Learning states assumptions, handles edge cases, explains trade-offs, and shows how to validate the result clearly.

  ExperianData ScientistMachine LearningTechnical Screen
• Clarify a BERT Keyphrase Span-Extraction TaskMedium

  Implement and clarify the post-processing decoder for a BERT-based span-extraction model that maps token-level start and end logits back to non-overlapping character-level keyphrases.

  FigmaMachine Learning EngineerMachine LearningTechnical Screen
• Build a robust ML pipelineMedium

  This question evaluates a candidate's competency in designing robust end-to-end ML pipelines, covering temporal data slicing, leakage controls, time-series cross-validation, feature store consistency, offline and business metrics, drift detection and monitoring, online rollout strategies, retraining triggers, and fairness assessment.

  Flatiron HealthData ScientistMachine LearningTechnical Screen
• Evaluate TPR/FPR, sigmoid, and activationsHard

  This question evaluates competency in binary classification metrics (confusion matrices, TPR/FPR), basic neural network forward computation (sigmoid neuron z and output), and knowledge of activation function selection and pitfalls.

  Goldman SachsMachine Learning EngineerMachine LearningOnline Assessment
• Address Overfitting in Supervised Learning ModelsMedium

  Evaluates supervised learning judgment around bias, variance, overfitting, and train-test performance gaps. Strong answers diagnose leakage, split errors, distribution shift, and high variance, then propose regularization, simpler models, validation, more data, and monitoring.

  GoogleData ScientistMachine LearningOnsite
• Build Classifier: Evaluate with AUROC for Imbalanced DataMedium

  Evaluates classifier design for detecting dead links in imbalanced URL data with leakage-safe validation. Strong answers cover leakage-safe features, grouped validation, AUROC versus accuracy, PR metrics, thresholding, deployment, and monitoring.

  GoogleData ScientistMachine LearningTechnical Screen
• Build Model to Predict Customer Contract RenewalMedium

  Evaluates renewal prediction modeling for enterprise Google Meet contracts. Strong answers define leakage-safe timelines, engineer usage, reliability, support, contract, and customer-health features, compare logistic regression with complex models, and evaluate calibration, lift, and business intervention value.

  GoogleData ScientistMachine LearningTechnical Screen
• Build and evaluate bad-link classifierMedium

  This question evaluates proficiency in applied machine learning classification, including feature design, training a logistic regression, handling severe class imbalance, selecting evaluation metrics and calibration, choosing thresholds under asymmetric costs, and planning offline-to-online validation and monitoring.

  GoogleData ScientistMachine LearningTechnical Screen
• Build and evaluate illegal-video classifierHard

  This question tests the ability to architect an end-to-end classification pipeline for high-stakes, multimodal content moderation while accounting for class imbalance, adversarial evasion, and operational safety.

  GoogleData ScientistMachine LearningTechnical Screen
• Compare Language-Model Post-Training MethodsHard

  This question tests your conceptual depth in LLM alignment by asking you to differentiate between supervised fine-tuning (SFT), reinforcement learning from human feedback (RLHF), direct preference optimization (DPO), and group-relative optimization (e.g., GRPO) based on their mathematical objectives and operational characteristics.

  GoogleMachine Learning EngineerMachine LearningTechnical Screen
• Compare Logistic Regression and Random Forest in Limited Data ScenariosMedium

  Evaluates binary classification model selection with limited labeled data. Strong answers explain logistic regression and cross-entropy loss, compare it with Random Forests, cover L1 and L2 regularization, diagnose overfitting, and contrast Random Forests with boosting.

  GoogleData ScientistMachine LearningOnsite
• Compare NLP tokenization and LLM recommendationsMedium

  This question tests the conceptual understanding of tokenization strategies in NLP and the architectural trade-offs when integrating LLMs into recommendation systems.

  GoogleMachine Learning EngineerMachine LearningOnsite
• Decide between two vendors under constraintsMedium

  This question evaluates a candidate's competency in decision modeling and cost‑sensitive machine learning, including causal inference for selection bias, propensity estimation, constrained optimization for capacity and SLA trade‑offs, and offline/online evaluation in the Machine Learning domain for a Data Scientist role.

  GoogleData ScientistMachine LearningOnsite
• Detect and Prevent Reward HackingHard

  Assess the candidate's understanding of reward hacking (specification gaming) in reinforcement learning from human feedback (RLHF), focusing on diagnosing exploitation patterns, designing detection metrics, applying conservative policy updates, and implementing recovery safeguards.

  GoogleMachine Learning EngineerMachine LearningTechnical Screen
• Diagnose and fix flawed model fitHard

  This question tests your ability to diagnose poor model performance in a binary classification task involving imbalanced data, high-cardinality categorical features, and potential bias.

  GoogleData ScientistMachine LearningTechnical Screen
• Explain GRPO-style training for diffusion modelsMedium

  This question tests your conceptual understanding of applying Group Relative Policy Optimization (GRPO) to diffusion models, focusing on how reinforcement learning can be used to steer image generation toward specific objective functions without explicit gradient feedback from the decoder.

  GoogleMachine Learning EngineerMachine LearningTechnical Screen
• Explain Importance Sampling in LLM TrainingHard

  Tests understanding of importance sampling in modern LLM fine-tuning, focusing on sequence-level likelihood ratios, exponential variance growth, support conditions, and practical mitigations like clipping and diagnostics.

  GoogleMachine Learning EngineerMachine LearningTechnical Screen
• Explain KL Divergence in Language ModelsHard

  Review KL divergence from its probability definition through directionality, support mismatch, sample estimates, and LLM regularization.

  GoogleMachine Learning EngineerMachine LearningTechnical Screen
• Explain LLM fine-tuning and generative modelsMedium

  This question tests your conceptual understanding of generative model architectures and the practical trade-offs involved in adapting large-scale language models to specific tasks.

  GoogleSoftware EngineerMachine LearningTechnical Screen
• Explain LLM lifecycle and trade-offsMedium

  This question evaluates a candidate's understanding of the end-to-end lifecycle of large language models, covering training data collection and filtering, pretraining objectives, transformer architecture, post-training methods like supervised fine-tuning and preference optimization, and common model or training variants.

  GoogleMachine Learning EngineerMachine LearningTechnical Screen
• Explain VAEs, ELBO, KL, and ReparameterizationHard

  Derive and explain the theoretical foundations of Variational Autoencoders (VAEs), linking the Evidence Lower Bound (ELBO) to network architecture, gradient estimation via the reparameterization trick, and posterior collapse.

  GoogleMachine Learning EngineerMachine LearningTechnical Screen
• Explain a favorite model end-to-endHard

  This question evaluates a candidate's end-to-end machine learning competencies, including problem framing, objective and loss selection, inductive biases, feature engineering, validation strategies, training and regularization practices, production issues, monitoring, and fairness/privacy considerations.

  GoogleData ScientistMachine LearningOnsite
• Explain linear regression to non‑technical stakeholdersMedium

  This question assesses the ability to communicate complex statistical models to non-technical stakeholders, focusing on clarity, intuition, and identifying limitations.

  GoogleData ScientistMachine LearningTechnical Screen