Machine Learning interview questions

Showing 481–510 of 572 results

• Compare Random Forests and Boosted Trees: Bias, Variance, SpeedMedium

  Evaluates practical trade-offs between Random Forests and gradient-boosted trees for tabular ML. Strong answers compare bias, variance, speed, interpretability, overfitting, production fit, and feature scaling needs.

  TikTokData ScientistMachine LearningTechnical Screen
• Define QKV for recommender cross-attentionHard

  This question evaluates understanding of Transformer-style cross-attention and the concrete design of Query, Key, and Value tensors for deep-learning recommender systems, testing representation semantics, embedding alignment, and interaction modeling between user history, candidate items, and context.

  TikTokMachine Learning EngineerMachine LearningTechnical Screen
• Design Real-Time Credit Card Fraud Detection SystemHard

  Evaluates real-time credit-card fraud detection system design under latency, drift, and asymmetric costs. Strong answers cover data, delayed labels, features, models, serving architecture, monitoring, retraining, and thresholding.

  TikTokData ScientistMachine LearningOnsite
• Design an ad-selection system across objectivesHard

  This question evaluates expertise in machine learning system design for ad selection and real-time bidding, covering predictive modeling and calibration, value unification across objectives, auction mechanics, budget pacing, exploration strategies, guardrails, and evaluation methodologies.

  TikTokData ScientistMachine LearningTechnical Screen
• Design multimodal deployment under compute limitsEasy

  This question evaluates competency in Machine Learning systems engineering for a Data Scientist role, covering multimodal model deployment and inference optimization, scalable retrieval (ANN and hybrid search), generalization and regularization concepts (overfitting, dropout), normalization methods, and RLHF, emphasizing both theoretical principles and engineering feasibility. It is commonly asked to assess reasoning about trade-offs between quality, latency, and cost in resource-constrained environments, validation via offline/online metrics, and the ability to bridge conceptual understanding with practical deployment and system-level design considerations.

  TikTokData ScientistMachine LearningOnsite
• Design recommendations objective balancing growth and monetizationHard

  This question evaluates a data scientist's competency in designing multi-objective recommender systems that balance user growth and creator monetization, covering ranking objective formulation, per-impression value estimation, constraint enforcement (latency, diversity, fairness/exposure), offline-to-online metric validation, cold-start and exploration strategies, and interference/attribution challenges within the Machine Learning domain. It is commonly asked to assess the ability to reason about trade-offs between business and user metrics, to define measurable objectives and validation plans, and to demonstrate both conceptual understanding and practical application of system design and evaluation.

  TikTokData ScientistMachine LearningHR Screen
• Estimate heterogeneous treatment effects with causal MLHard

  This question evaluates a candidate's competency in causal machine learning, heterogeneous treatment effect estimation, off-policy evaluation, and fairness-aware policy design using observational data.

  TikTokData ScientistMachine LearningTechnical Screen
• Explain DPO and construct its training dataMedium

  This question evaluates understanding of Direct Preference Optimization (DPO) for fine-tuning large language models, assessing conceptual differences from PPO-based RLHF and the competency to design pairwise preference training datasets.

  TikTokSoftware EngineerMachine LearningTechnical Screen
• Explain FlashAttention, KV cache, and RoPEMedium

  This problem tests your understanding of core transformer efficiency techniques, specifically focusing on how they improve memory footprint, latency, and sequence length handling.

  TikTokMachine Learning EngineerMachine LearningTechnical Screen
• Explain and tune XGBoost; prevent overfittingHard

  This question evaluates understanding of XGBoost tree-boosting internals, hyperparameter impacts on bias/variance/training time, strategies for imbalanced binary classification and leakage-aware validation, post-training calibration, and interpretability for tabular fraud detection.

  TikTokData ScientistMachine LearningTechnical Screen
• Explain your VLM project end-to-endMedium

  This question evaluates proficiency in vision-language model engineering, including model architecture (vision encoder, language model, fusion), data curation and distribution, training recipes, evaluation metrics, inference latency, and limitations analysis within the Machine Learning domain of multimodal/Vision-Language models.

  TikTokMachine Learning EngineerMachine LearningTechnical Screen
• How do you choose a classification threshold?Easy

  This question evaluates a data scientist's competency in binary classification threshold selection, end-to-end ML pipeline design, model evaluation, and operational trade-offs including class imbalance, asymmetric error costs, and limited escalation capacity, within the Machine Learning domain of model evaluation and deployment.

  TikTokData ScientistMachine LearningTechnical Screen
• When prioritize precision vs recallEasy

  This question evaluates a data scientist's competency with binary classification metrics (precision and recall), cost-sensitive decision-making, and operating-threshold selection for productized models in the Machine Learning domain.

  TikTokData ScientistMachine LearningTechnical Screen
• Explain Why Constant Neural-Network Initialization FailsHard

  This question tests the conceptual understanding of weight initialization strategies and the symmetry-breaking property required for neural network gradient-based optimization.

  TowerData ScientistMachine LearningTechnical Screen
• Explain the KV Cache and Prefill-Decode Disaggregation in LLM ServingMedium

  Explain how the KV cache speeds up autoregressive decoding in transformer language models, estimate how much GPU memory it needs, and describe prefill-decode disaggregation in LLM serving. It tests understanding of compute-bound versus memory-bound inference phases, cache sizing and the trade-offs of splitting the phases across GPU pools.

  TowerMachine Learning EngineerMachine LearningTechnical Screen
• Explain ML basics and recommender tuningMedium

  This question evaluates core machine learning competencies including overfitting and regularization, ensemble methods like bagging, linear and logistic regression, transformer architectures, optimizer trade-offs (SGD vs Adam), hyperparameter tuning in production, and end-to-end recommender system design and deployment.

  TubiMachine Learning EngineerMachine LearningTechnical Screen
• Machine Learning Fundamentals: Tree Models, Training, Evaluation, and EmbeddingsMedium

  This question assesses foundational machine learning knowledge across multiple core domains, including tree-based models, supervised training, model evaluation, embeddings, and transformer architecture. It is commonly used in ML engineer interviews to gauge whether a candidate can explain key concepts with correct mechanics, articulate trade-offs, and reason through failure modes.

  TubiMachine Learning EngineerMachine LearningTechnical Screen
• Estimate Volatility from Gaussian and Brownian SamplesMedium

  This question evaluates skills in statistical estimation and stochastic process modeling, specifically parameter estimation for Gaussian samples and volatility estimation from discrete Brownian motion observations, focusing on estimating the volatility parameter (sigma) or its variance (sigma^2).

  TudorData ScientistMachine LearningTechnical Screen
• Compare XGBoost and LightGBMHard

  This question evaluates a data scientist's understanding of gradient-boosted tree implementations and practical model trade-offs, covering competencies in algorithmic differences, performance and memory characteristics, handling of missing and categorical features, regularization risks, and hyperparameter considerations for production models.

  TuroData ScientistMachine LearningOnsite
• Design real-time live-stream recommendationsHard

  This question evaluates competency in designing real-time recommender systems within the Machine Learning domain, covering objectives and label construction for streaming events, feature engineering across user/creator/stream/context, model and loss choices, bias mitigation, evaluation metrics, and low-latency serving and freshness constraints.

  TwitchData ScientistMachine LearningTechnical Screen
• Analyze Temperatures and Update RegressionMedium

  This question evaluates proficiency in time-series and statistical analysis (defining volatility and similarity metrics), supervised regression modeling with mean squared error evaluation, greedy feature selection, and streaming/no-intercept linear regression updates.

  Two SigmaData ScientistMachine LearningOnline Assessment
• Build a Housing Price Prediction Model from Historical DataHard

  This question evaluates a data scientist's competency in end-to-end supervised machine learning for real-estate pricing, focusing on precise outcome definition, feature engineering, handling missing and time-varying data, and robustness to market regime changes.

  Two SigmaData ScientistMachine LearningOnsite
• Daily Temperatures by Town: Linear Models to Predict NYC TemperatureMedium

  This question evaluates proficiency in ordinary least squares linear regression (with intercept), basic statistical summaries such as standard deviation and median, and model-evaluation metrics like in-sample mean squared error applied to aligned daily time-series temperature data.

  Two SigmaQuantitative ResearcherMachine LearningTechnical Screen
• Derive correlation bounds and omitted-variable biasHard

  This problem tests your understanding of the constraints on correlation matrices and your ability to analytically derive the bias introduced in linear models when a relevant predictor is ignored.

  Two SigmaData ScientistMachine LearningTechnical Screen
• Design a house-price prediction modelEasy

  This problem evaluates your ability to design a robust end-to-end machine learning pipeline for house price prediction, focusing on data preprocessing, feature engineering, and validation strategy.

  Two SigmaData ScientistMachine LearningTechnical Screen
• Design features for house price predictionMedium

  This question tests your ability to engineer features from heterogeneous tabular data, handle missing values, and select appropriate evaluation metrics for a high-stakes regression problem.

  Two SigmaData ScientistMachine LearningTechnical Screen
• Forecast bikes available at a stationHard

  This question tests the ability to design a robust time-series forecasting pipeline, specifically addressing feature engineering for cyclic temporal patterns and the selection of appropriate validation strategies for sequential data.

  Two SigmaData ScientistMachine LearningTechnical Screen
• How detect duplicate card records?Medium

  This question evaluates a candidate's competency in duplicate detection, record linkage, entity resolution, and data quality assessment within credit card transaction datasets.

  Two SigmaData ScientistMachine LearningTechnical Screen
• How to forecast bike dock demandEasy

  This problem tests your ability to design a predictive pipeline for spatio-temporal demand, focusing on feature extraction from time-series data and robust validation strategies.

  Two SigmaData ScientistMachine LearningTechnical Screen
• How would you forecast bike demand?Hard

  This problem tests your ability to design a robust machine learning pipeline for time-series demand forecasting, focusing on feature engineering for exogenous variables and evaluation metrics for spatial-temporal data.

  Two SigmaData ScientistMachine LearningTechnical Screen