Applied Scientist Interview Questions

Showing 1–14 of 14 results

• AmazonEasy·Applied Scientist
  Describe overfitting and L1/L2 regularization

  This question evaluates understanding of model generalization and regularization techniques by probing knowledge of overfitting and L1/L2 regularization and their effects on model parameters.

  Machine LearningTechnical Screen
• AmazonHard·Applied Scientist
  Explain Core ML Interview Concepts

  This question evaluates core machine learning fundamentals including statistical modeling assumptions and loss functions (linear and logistic regression), ensemble methods and feature sampling in random forests, optimization algorithms (Adam versus stochastic gradient descent), and neural network capacity and training dynamics.

  Machine LearningTechnical Screen
• AmazonMedium·Applied Scientist
  Explain LLM architecture, tuning, evaluation

  This question evaluates understanding of transformer-based LLM architectures, positional embedding variants, parameter-efficient fine-tuning (PEFT) approaches, regularization strategies, and evaluation methodologies within the Machine Learning domain focused on large language models and NLP, emphasizing both conceptual understanding and practical application. It is commonly asked to assess the interviewee’s ability to reason about architecture and tuning trade-offs, model generalization and stability, and the design of reliable offline and online evaluation pipelines for production-minded language model systems.

  Machine LearningTechnical Screen
• AmazonMedium·Applied Scientist
  Explain core ML concepts and diagnostics

  This question evaluates mastery of core Machine Learning concepts and diagnostics, covering statistical inference (p-values), overfitting/underfitting and bias-variance, causal inference approaches, encoding/decoding, optimization and backpropagation, gradient stability, handling imbalanced data, evaluation metrics versus accuracy, and experimental A/B testing. It is commonly asked in ML interviews to assess breadth and depth in the Machine Learning domain and probes both conceptual understanding and practical application by testing theoretical knowledge, recognition of common failure modes, and diagnostic reasoning used to validate models and experiments.

  Machine LearningTechnical Screen
• AmazonHard·Applied Scientist
  Explain imbalance, metrics, bias-variance, Transformers vs. CNNs

  An Amazon machine-learning-engineer technical screen covering class-imbalance handling at the data, loss, and decision levels; leakage-aware validation; the bias–variance tradeoff; metric selection for rare-positive fraud data (PR-AUC, precision@k, calibration); and Transformer vs. CNN inductive biases. Includes worked numeric examples for cost-based thresholds and metric interpretation.

  Machine LearningTechnical Screen
• AmazonMedium·Applied Scientist
  Explain surprisal and its units

  This question evaluates understanding of probabilistic information measures in language modeling, specifically the concept of surprisal, its units, and its relationship to evaluation metrics such as cross-entropy and perplexity within the Machine Learning/NLP domain.

  Machine LearningTechnical Screen
• AmazonHard·Applied Scientist
  GRPO Deep Dive: Critic-Free RL, Parallelism, MLA, and Reward Design for a Reasoning LLM

  This question tests your deep understanding of Group Relative Policy Optimization (GRPO) for post-training reasoning models, including how critic-free optimization alters training dynamics, distributed memory scaling, KV-cache efficiency via Multi-Head Latent Attention (MLA), and reward hacking mitigations.

  Machine LearningTechnical ScreenReinforcement LearningLLM Post-TrainingGRPO
• AmazonMedium·Applied Scientist
  Implement SGD for linear regression and derive gradients

  This question evaluates understanding of gradient-based optimization and linear model training, including competency in deriving gradients for mean squared error and implementing stochastic (or mini-batch) gradient descent for parameter estimation.

  Machine LearningTechnical Screen
• AmazonEasy·Applied Scientist
  List hyperparameter tuning methods

  This question evaluates understanding of hyperparameter tuning methods, including the competencies to compare different search strategies, analyze trade-offs in efficiency and parallelization, and design experiments to optimize model performance.

  Machine LearningTechnical Screen
• BByteDanceMedium·Applied Scientist
  Explain GRPO for LLM Reinforcement Learning and How It Differs from PPO

  An ML question asking you to explain Group Relative Policy Optimization (GRPO) for fine-tuning large language models with reinforcement learning. It tests your grasp of group-normalized advantages, the clipped objective with a KL penalty, the contrast with PPO and its value model, and common failure modes.

  Machine LearningTechnical Screen
• BByteDanceMedium·Applied Scientist
  Implement Multi-Head Attention from Scratch

  An ML coding question asking you to implement multi-head attention from scratch in NumPy or PyTorch, from the query, key and value projections to the output projection. It tests tensor shape handling, scaled dot-product attention, masking, numerical stability and complexity analysis.

  Machine LearningTechnical Screen
• BByteDanceMedium·Applied Scientist
  Write the Cross-Entropy Loss and Compute Precision, Recall and F1

  An ML fundamentals question in two parts: write the cross-entropy loss for binary and multi-class classification and explain why it is used, then compute precision, recall, F1 and accuracy for a described imbalanced classifier. It tests exact formulas, careful arithmetic and metric choice.

  Machine LearningTechnical Screen
• EeBayEasy·Applied Scientist
  Named entity recognition: framing, algorithm choice, data processing, and metrics

  A machine learning question on named entity recognition for short e-commerce text such as listing titles and search queries. It asks you to define the task, compare NER algorithm families and choose one, plan labeling and data processing, and select entity-level and product-level metrics.

  Machine LearningTechnical Screen
• EeBayEasy·Applied Scientist
  Optimizing BERT inference latency and throughput for a production NLP service

  An ML engineering question on speeding up a fine-tuned BERT encoder that has become too slow or too costly to serve in production. It tests profiling, model compression, runtime and serving optimizations such as batching and padding control, and how to verify that accuracy does not regress.

  Machine LearningTechnical Screen