practice a little every day, paws crossed ♡
Software Engineer Interview Questions
Practice 9403 coding, system design and machine learning questions from 546 companies, sorted by role, topic and interview round. Every question has step-by-step hints, a code runner that supports 8 languages and an AI mentor; members also get the full solution walkthrough.
- 9403
- Questions
- 546
- Companies
- 11
- Topics
- 8
- Languages
Showing 1–30 of 53 results
- AMDMediumSoftware EngineerExplain the ReAct Agent Loop
Explain a ReAct tool loop through state, action selection, observations, recovery, stopping criteria, and validation of task completion.
Machine LearningOnsite - AmazonMediumSoftware EngineerEvaluating Generative AI Use and Accuracy
Tests a candidate's ability to assess when Generative AI is the right solution, choose appropriate architecture patterns (such as RAG or fine-tuning), and rigorously measure and improve model output accuracy in production.
Machine LearningOnsiteGenerative AIRAGML Evaluation - AmazonMediumSoftware EngineerExplain Decoder-Only LLM Architecture and How vLLM Serves It Efficiently
A two-part conversational question for AI infrastructure roles: explain how a decoder-only transformer turns a prompt into next-token probabilities, then explain how vLLM serves such models efficiently. It tests transformer internals, KV cache sizing, prefill versus decode bottlenecks, paged KV memory and continuous batching.
Machine LearningOnsite - AmazonMediumSoftware EngineerExplain overfitting, regularization, and LLM techniques
This question tests the conceptual understanding of model generalization trade-offs, the mathematical mechanics of regularization, and the practical application of modern LLM architectures.
Machine LearningTechnical ScreenRegularizationLLMRAG - AmazonHardSoftware EngineerUse a Fitted Line to Predict a Future Data Point
This question tests the ability to implement a simple linear regression predictor and analyze the reliability of extrapolating time-series data using basic statistical heuristics.
Machine LearningTechnical ScreenRegressionTime SeriesStatistics - AncestryMediumSoftware EngineerExplain learning paradigms, loss, and embeddings
This question evaluates understanding of core machine learning concepts—learning paradigms (supervised, unsupervised, self-supervised), loss functions, and embeddings—and the competency to explain their roles and interrelations in model training and inference.
Machine LearningTechnical Screen - AnthropicMediumSoftware EngineerDebug a GRPO training loop and explain ratios
This question evaluates debugging and implementation knowledge for on-policy reinforcement learning, focusing on GRPO/PPO-style training loops, importance-sampling ratios, log-prob computations, masking, and advantage normalization.
Machine LearningTechnical Screen - AnthropicMediumSoftware EngineerIdentify the Most Important AI Safety Risk
Tests your ability to analyze AI alignment failure modes—specifically sycophancy and emotional over-reliance—and formulate rigorous evaluation metrics, training mitigations, and product-level guardrails.
Machine LearningHR ScreenAI SafetyRLHFSycophancy - AnthropicHardSoftware EngineerImplement and analyze custom attention
This question evaluates implementation and analysis skills for scaled dot-product attention, testing competencies in efficient tensorized PyTorch coding, numerical stability and mixed-precision handling, masking semantics (causal and padding), multi-head shape correctness, and unit-test validation including gradient and edge-case behavior.
Machine LearningOnsite - AnthropicMediumSoftware EngineerImplement and derive backprop from scratch
This question evaluates understanding and practical implementation of neural network fundamentals, specifically analytic backpropagation and gradient derivation, numerically stable binary cross-entropy computation, parameter initialization, gradient descent updates, and gradient checking.
Machine LearningOnsite - Axq CapitalMediumSoftware EngineerExplain What a Neural Network Is and How It Learns
Explain what a neural network is: how units and layers turn weighted inputs into outputs, why non-linear activations are essential, and how the weights are learned with a loss function, backpropagation, and gradient descent. It tests core machine learning fundamentals, including overfitting and when simpler models are the better choice.
Machine LearningHR Screen - ByteDanceMediumSoftware EngineerChoose Precision, Recall, or F1 for Agent Evaluation
This question tests the ability to map business objectives and error costs to appropriate classification metrics in a real-world system context.
Machine LearningTechnical ScreenMetricsClassificationModel Evaluation - ByteDanceMediumSoftware EngineerHand-Write a Multi-Head Attention Layer: Projections, Head Splitting and Masking
Write a multi-head attention layer from scratch: project queries, keys and values, split them into heads, apply scaled dot-product attention with optional padding or causal masks, and merge the heads. It tests tensor shape discipline, numerical stability, and the cost of attention as sequence length grows.
Machine LearningOnsite - C3 AIMediumSoftware EngineerEvaluate Factual Accuracy and Consistency in a Domain-Specific RAG Chatbot
Evaluate a rare-disease RAG chatbot using domain consistency and expert review, and distinguish factual accuracy from retrieval metrics and lexical overlap.
Machine LearningOnline Assessment - CrestaMediumSoftware EngineerDesign sequence decoding with greedy and beam search
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 Design sequence decoding with greedy and beam search states assumptions, handles edge cases, explains trade-offs, and shows how to validate the result clearly.
Machine LearningTechnical Screen - DeshawHardSoftware EngineerExplain and Compare SGD and Adam
Explain how stochastic gradient descent updates model parameters and how Adam modifies those updates using moving estimates of first and second gradient moments. Cover data and labels, leakage-safe features, baselines and model choice, offline evaluation, deployment constraints, monitoring, and drift.
Machine LearningTechnical Screen - GoogleMediumSoftware EngineerExplain LLM fine-tuning and generative models
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.
Machine LearningTechnical ScreenLLMFine-tuningGenerative Models - GoogleMediumSoftware EngineerHow Would You Verify a Smiling-Face Detection Feature in Google Photos?
A role-related question for software engineers: if your feature detects smiling faces in Google Photos, how would you verify that the detection is correct? It tests defining ground truth, choosing evaluation metrics and thresholds, covering hard cases and demographic slices, and validating the feature before and after launch.
Machine LearningTechnical Screen - HarveyMediumSoftware EngineerEmbedding Retrieval with Cosine Similarity in a Notebook, and Choosing the Metric
Implement embedding-based document retrieval with cosine similarity in a Python notebook, then explain how cosine similarity relates to dot product and Euclidean distance and how to measure retrieval quality. It tests vectorized top-k search, deterministic ties, index metric choice and evaluation metrics.
Machine LearningOnsite - Hudson River TradingHardSoftware EngineerBuild a Heart Disease Baseline
This question evaluates proficiency in applied machine learning workflows, including data inspection and cleaning, exploratory data analysis, feature engineering, baseline binary classification modeling, and evaluation using Python libraries like pandas and seaborn.
Machine LearningOnsite - IntuitMediumSoftware EngineerExplain LLMs, AI-native vs. conventional apps, daily AI use, and the dangers of AI
Staff-level backend interview discussion on AI fluency. Explain what a large language model is, contrast AI-native products with conventional apps that add AI features, describe how you use AI tools in daily engineering work, and weigh the main dangers of AI with concrete mitigations.
Machine LearningTechnical Screen - IntuitEasySoftware EngineerWhen should products use AI?
This question evaluates product-oriented decision-making about AI adoption, testing competencies in machine learning product strategy, trade-off analysis (accuracy, latency, cost, explainability, safety, and user trust), and the ability to distinguish AI-native architectures from conventional feature additions.
Machine LearningTechnical Screen - Jump TradingHardSoftware EngineerAccelerate Tree-Model Inference
Design an end-to-end strategy to accelerate inference latency and throughput for an existing gradient-boosted decision tree (GBDT) model in a low-latency trading environment while preserving exact prediction fidelity. Address feature engineering constraints, low-level execution optimizations, verification of numerical equivalence, and production drift monitoring.
Machine LearningTechnical ScreenTree ModelsInference OptimizationLow-Latency Systems - MicrosoftMediumSoftware EngineerGreedy and Top-k Beam Search Decoding That Returns the Sentence Probability
A coding question on decoding a sequence from a next-token probability distribution: implement greedy decoding, extend it to beam search that keeps the top k partial sequences, and return the chosen sequence with its overall probability. It also probes why long sequences tend to receive low probability and how decoding should account for that.
Machine LearningTechnical Screen - MillenniumMediumSoftware EngineerHow would you model stock price prediction?
This problem tests the candidate's ability to design an end-to-end machine learning pipeline for financial time-series forecasting, specifically focusing on the nuances of feature engineering and evaluation in a non-stationary environment.
Machine LearningTechnical ScreenTime-SeriesFeature EngineeringFinancial Engineering - Mistral AIMediumSoftware EngineerCompare RMSNorm and LayerNorm in Transformer Blocks
Compare LayerNorm and RMSNorm equations, centering, shift behavior, hidden-feature axes, numerical precision, and the tradeoffs behind simpler normalization.
Machine LearningTechnical Screen - Mistral AIMediumSoftware EngineerCompare Reward-Model RLHF and Direct Preference Optimization
Compare reward-model RLHF and DPO through preference data, policy objectives, reference regularization, training stages, and evaluation under distribution shift.
Machine LearningTechnical Screen - Mistral AIMediumSoftware EngineerCompare Tensor Parallelism and Pipeline Parallelism
Compare transformer tensor and pipeline parallelism through parameter partitioning, collectives, activation transfers, microbatch bubbles, memory, and inference latency.
Machine LearningTechnical Screen - Mistral AIMediumSoftware EngineerExplain Adam, AdamW, and Decoupled Weight Decay
Understand Adam moment estimates and bias correction, then compare gradient-based L2 regularization with AdamW’s decoupled weight decay and learning-rate interaction.
Machine LearningTechnical Screen - Mistral AIMediumSoftware EngineerExplain Mixture of Experts and Expert Parallelism
Explain sparse mixture-of-experts routing and expert parallelism, including token dispatch, weighted combination, load balancing, overflow policies, and communication costs.
Machine LearningTechnical Screen
Amazon
Anthropic
C3 AI
Cresta
Google
Harvey
Hudson River Trading
Intuit
Microsoft
Millennium
Mistral AI