Applied Scientist Interview Questions

Showing 1–30 of 41 results

• AmazonHard·Applied Scientist
  Build an end-to-end ML pipeline

  Build an end-to-end ML pipeline evaluates ML product requirements, data/labeling, modeling, serving architecture, evaluation, monitoring, and trade-offs in a realistic interview setting. A strong answer states assumptions, handles edge cases, explains trade-offs, and shows how to validate the result clearly.

  ML System DesignOnsite
• AmazonMedium·Applied Scientist
  Compute and Explain Mean and Variance

  This question tests your understanding of statistical estimators for mean and variance, the derivation of Bessel's correction, catastrophic cancellation in naive computations, and numerically stable single-pass algorithms like Welford's.

  Statistics & MathOnsiteStatisticsNumerical AnalysisAlgorithms
• AmazonHard·Applied Scientist
  Deep-dive your GenAI project architecture

  Deep-dive your GenAI project architecture evaluates ML product requirements, data/labeling, modeling, serving architecture, evaluation, monitoring, and trade-offs in a realistic interview setting. A strong answer states assumptions, handles edge cases, explains trade-offs, and shows how to validate the result clearly.

  ML System DesignOnsite
• AmazonMedium·Applied Scientist
  Describe a decision with incomplete information

  Describe a decision with incomplete information evaluates behavioral evidence, ownership, communication, trade-offs, and measurable outcomes in a realistic interview setting. A strong answer states assumptions, handles edge cases, explains trade-offs, and shows how to validate the result clearly.

  Behavioral & LeadershipOnsite
• AmazonMedium·Applied Scientist
  Design LFU cache with distributed extension

  This question evaluates understanding of cache design and frequency-based eviction policies, efficient data structures that support O(1) operations, and techniques for approximate frequency estimation and summarization in distributed streaming environments.

  Coding & AlgorithmsOnsite
• AmazonHard·Applied Scientist
  Design a Multimodal Neural Network

  This question evaluates a candidate's competency in designing production-grade multimodal machine learning systems, including architecture choices for text and image encoders, cross-modal fusion strategies, training objectives for joint retrieval and classification, robustness to missing or noisy modalities, and considerations for scalability and low-latency serving. Commonly asked in ML system design interviews to assess both conceptual understanding and practical application of machine learning engineering principles, it falls under the ML System Design domain and probes abilities in model alignment, evaluation metrics, deployment trade-offs, and domain adaptation strategies.

  ML System DesignOnsite
• AmazonMedium·Applied Scientist
  Design a computer-use agent end-to-end

  This problem evaluates your ability to design a multimodal agent capable of interpreting UI elements and executing sequences of actions to complete user-defined tasks on a computer screen.

  ML System DesignOnsiteVLMReinforcement LearningMultimodal
• AmazonMedium·Applied Scientist
  Design an LLM quality validation system

  This question evaluates an engineer's competency in ML system design for validating large language models, testing knowledge of evaluation metrics, validation scopes from internal numerics to final responses, system architecture for automated validation pipelines, and operational concerns within the ML System Design domain; it requires both conceptual architecture understanding and practical application-level thinking. It is commonly asked to assess the ability to reason about correctness, safety, regression risk, scalability, reproducibility, and monitoring trade-offs when deploying production-grade validation pipelines and to gauge familiarity with complementary evaluation methods and end-to-end data flows.

  ML System DesignOnsite
• AmazonMedium·Applied Scientist
  Design an S3-like object storage service

  This question evaluates understanding of distributed systems, object storage architecture, REST API design, multipart upload semantics, replication and durability strategies, failure recovery, and scalability concerns.

  System DesignOnsite
• AmazonHard·Applied Scientist
  Design systems for global request detection and labeling

  This problem tests the architecture of a high-throughput streaming system capable of detecting anomalies in real-time, managing extreme class imbalance for labeling, and maintaining low-latency inference at a global scale.

  ML System DesignOnsiteStreamingReal-time InferenceImbalanced Learning
• AmazonMedium·Applied Scientist
  Designing a Bayesian Classifier for Case Analysis

  Tests your ability to design a probabilistic Bayesian classification system under conditions of informative missingness, non-stationary class priors, and strict calibration requirements.

  Machine LearningOnsiteBayesian ModelingMissing DataModel Calibration
• AmazonHard·Applied Scientist
  Explain Collaborative Filtering Approaches

  This question evaluates understanding of collaborative filtering approaches and related competencies such as neighborhood-based versus matrix factorization methods, differences for explicit versus implicit feedback, pointwise versus pairwise loss trade-offs, regularization, cold-start and popularity-bias mitigation, evaluation metrics, and scalability considerations. It is asked in the Machine Learning domain to assess both conceptual understanding and practical application for designing, evaluating (e.g., NDCG, MAP) and scaling recommender systems in technical interviews.

  Machine LearningOnsite
• AmazonMedium·Applied Scientist
  Explain Layer Normalization in Transformers

  This question evaluates a candidate's understanding of normalization techniques in deep learning—specifically Layer Normalization in Transformer blocks—covering the LayerNorm equation and the roles of gamma, beta, and epsilon, as well as contrasts with BatchNorm and RMSNorm, in the Machine Learning domain.

  Machine LearningOnsite
• AmazonHard·Applied Scientist
  Explain Logistic Regression Fundamentals

  This question evaluates understanding of logistic regression fundamentals, covering probabilistic modeling, Bernoulli likelihood and logit-link derivation, loss and gradient characterization, calibration and miscalibration remedies, regularization trade-offs, feature scaling and interactions, class imbalance strategies, evaluation metrics for skewed data, coefficient interpretation, and common failure modes. It is commonly asked in technical interviews to assess both theoretical and practical competencies in supervised learning within the Machine Learning domain, with a level of abstraction spanning conceptual derivation and practical application of model behavior and evaluation.

  Machine LearningOnsite
• AmazonMedium·Applied Scientist
  Explain ML evaluation, sequence models, and optimizers

  This question evaluates a candidate's competency in machine learning model evaluation and metrics, sequence modeling trade-offs between transformers and RNNs, image distributional shift detection, and comparative understanding of optimization algorithms within the Machine Learning domain.

  Machine LearningOnsite
• AmazonHard·Applied Scientist
  Explain NLP/RL concepts used in LLM agents

  This question tests the ability to synthesize concepts across LLM agent architecture, RAG systems, and reinforcement learning, focusing on architectural design trade-offs and decision-making logic.

  Machine LearningOnsiteLLM-AgentsRAGReinforcement-Learning
• AmazonMedium·Applied Scientist
  Explain XGBoost Parallelism Strategies

  This question evaluates a candidate's understanding of parallelism and system-level performance in gradient-boosted tree implementations, covering concepts such as histogram-based split finding, sparse feature handling, cache-friendly data layouts, thread-level work partitioning, and multi-machine data-parallel synchronization.

  Machine LearningOnsite
• AmazonHard·Applied Scientist
  Explain key ML theory and techniques

  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 key ML theory and techniques states assumptions, handles edge cases, explains trade-offs, and shows how to validate the result clearly.

  Machine LearningOnsite
• AmazonMedium·Applied Scientist
  Explain parallelism and collectives in training

  This question evaluates a candidate's competency in designing scalable distributed training systems, covering parallelism strategies (data, model/tensor, pipeline), communication collectives (all-reduce, all-gather, reduce-scatter, broadcast), and tensor-level layer partitioning such as column- and row-parallel splits.

  ML System DesignOnsite
• AmazonMedium·Applied Scientist
  Explain why CTR rises but CVR unchanged

  This question tests the ability to diagnose Simpson's Paradox or composition shifts in funnel metrics when engagement increases without a corresponding lift in final conversion.

  Analytics & ExperimentationOnsiteAnalyticsExperimentationFunnel Analysis
• AmazonMedium·Applied Scientist
  Find shortest path in a grid with obstacles

  This question evaluates understanding of pathfinding in grid-based environments, graph traversal concepts, and algorithmic complexity when navigating obstacles and boundary constraints in an unweighted 2D search space.

  Coding & AlgorithmsOnsite
• AmazonMedium·Applied Scientist
  From Early NLP Embeddings to Modern LLM Embeddings

  Evaluate the architectural and practical evolution of NLP text embeddings, and formulate a decision framework comparing small bi-encoders versus large LLM-based embedding models based on retrieval quality, latency, and operational cost.

  Machine LearningOnsiteEmbeddingsNLPInformation Retrieval
• AmazonMedium·Applied Scientist
  Implement K-means and solve interval/frequency tasks

  This multi-part question evaluates understanding of K-means clustering, interval-merging algorithms, and frequency-based top-k retrieval, assessing competencies in unsupervised learning concepts, algorithmic problem solving, data structures, and complexity analysis.

  Coding & AlgorithmsOnsite
• AmazonMedium·Applied Scientist
  Implement Multi-Head Attention from Scratch in NumPy

  This problem tests your understanding of Transformer components by requiring an implementation of Multi-Head Scaled Dot-Product Attention using only NumPy, emphasizing matrix manipulations and numerical stability.

  Coding & AlgorithmsOnsiteTransformersNumPyAttention
• AmazonMedium·Applied Scientist
  Implement PyTorch training loop

  This question evaluates practical implementation skills in PyTorch, focusing on model and device management, batch-wise training mechanics, gradient handling, and optimizer interaction.

  Coding & AlgorithmsOnsite
• AmazonMedium·Applied Scientist
  Implement Top-p (Nucleus) Sampling in NumPy

  This problem tests your ability to implement the nucleus sampling algorithm from scratch using NumPy, focusing on probability distribution processing and efficient array masking.

  Coding & AlgorithmsOnsiteNumPyNLPProbability
• AmazonMedium·Applied Scientist
  Implement decoder-only GPT-style transformer

  This question evaluates implementation-level mastery of transformer architectures, including multi-head masked self-attention, position-wise feed-forward networks, residual connections, and precise tensor-shape reasoning for a decoder-only language model, within the Coding & Algorithms domain of deep learning and neural network engineering.

  Coding & AlgorithmsOnsite
• AmazonMedium·Applied Scientist
  Implement integer division without using division

  This question evaluates proficiency in implementing low-level integer arithmetic and algorithmic optimization, including correct handling of overflow and edge cases within 32-bit signed integer constraints.

  Coding & AlgorithmsOnsite
• AmazonMedium·Applied Scientist
  Implement inventory allocation with backorders

  Implement inventory allocation with backorders evaluates algorithm design, data structures, correctness, complexity, edge cases, and implementation details in a realistic interview setting. A strong answer states assumptions, handles edge cases, explains trade-offs, and shows how to validate the result clearly.

  Coding & AlgorithmsOnsite
• AmazonMedium·Applied Scientist
  LLM Fundamentals: Tokenization Design and KL-Regularized SFT

  This problem tests your understanding of subword tokenization strategies and the theoretical motivation behind KL-divergence constraints during supervised fine-tuning.

  Machine LearningOnsiteLLMTokenizationSFT