practice a little every day, paws crossed ♡
Machine Learning 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 91–120 of 351 results
- FigmaMediumMachine Learning EngineerImplement Layer History and Grid Counting
This question tests state management through a Command pattern implementation for undo/redo operations and spatial reasoning using depth-first search (DFS) to identify connected components in a 2D grid.
Coding & AlgorithmsOnsiteDesign PatternsStackGraphs - FireworksMediumMachine Learning EngineerRecommend Connections Through Mutual Friends and Evaluate the Ranking
Design a connection recommendation system based on mutual connections, define bounded graph affinity scores, and evaluate the ranking pipeline using metrics like NDCG with independent labels.
ML System DesignOnsiteGraph MLRecommender SystemsRanking - FurtheraiMediumMachine Learning EngineerBuild a Create/Execute Workflow Service With an AI Assistant: Plan, Implement, Verify
A 45-minute practical coding exercise: build a small service with create-workflow and execute-workflow APIs while using an AI coding assistant. It evaluates planning, implementing and verifying with the assistant, along with validation of workflow definitions, dependency-ordered execution and step failure handling.
Software Engineering FundamentalsOnsite - FurtheraiMediumMachine Learning EngineerCutting Scope to Ship the Core Customer Problem First: Working Pragmatically With Product
A behavioral question about working pragmatically with a product team: describe a real time you cut scope to ship a minimal first version that solved the customer's core problem, and explain how you and a product manager decide what goes into a first release, what waits, and how disagreements are resolved.
Behavioral & LeadershipOnsite - FurtheraiMediumMachine Learning EngineerMap an Insurance Underwriting Workflow and Pick What AI Automation Should Fix First
A business-curiosity question for an engineer at a company automating insurance underwriting: describe the underwriting workflow from submission to decision, decide which pain points AI should automate first and how to measure the result, and ask questions that show genuine interest in the customer's work.
Product Design & StrategyOnsite - FurtheraiMediumMachine Learning EngineerMerge Overlapping Closed Intervals Into a Sorted Disjoint List
A coding question that merges an unsorted list of closed integer intervals into disjoint intervals sorted by start. Intervals that overlap or merely touch at an endpoint must be combined, so it tests exact endpoint semantics, transitive merges, and nested or duplicate intervals.
Coding & AlgorithmsOnsite - FurtheraiMediumMachine Learning EngineerMotivation for an Early-Stage AI Startup, Hands-On AI Experience, and Startup Fit
A fit conversation for a staff applied AI engineering role at an early-stage startup: explain your relevant projects and hands-on experience with AI, why you are looking now, why an early-stage company, and what startup experience you bring, as probed by a hiring manager and a CTO.
Behavioral & LeadershipOnsite - FurtheraiMediumMachine Learning EngineerWalk Through an AI Project You Built: Decisions, Alternatives, Trade-offs, Impact
A behavioral deep-dive question asking a senior candidate to walk through a project where AI was central, covering the problem, the technical difficulty, architecture decisions, alternatives considered, trade-offs and lessons, and measured business or customer impact, with close probing of the alternatives and trade-offs.
Behavioral & LeadershipOnsite - GleanMediumMachine Learning EngineerAI-Paired Rate-Limited Wikipedia Crawler That Prioritizes Unseen Title Initials
Pair with an AI coding assistant to build a rate-limited Wikipedia crawler that first crawls pages whose title initial has not yet been covered, then crawls randomly once all 26 letters are done. Tests frontier design, link normalization, rate limiting and disciplined use of AI-generated code.
Software Engineering FundamentalsOnsite - GleanMediumMachine Learning EngineerDecide If a Passenger Can Reach a Station Using Time-Indexed Train Routes and Transfers
Given trains described as time-indexed lists of stations, decide whether a passenger starting at one station at time zero can reach another by riding, waiting and transferring. Tests modeling station and time as search states, handling waits and same-time transfers, and breadth-first search over a time-expanded graph.
Coding & AlgorithmsOnsite - GleanMediumMachine Learning EngineerDesign Employee-to-Employee Distance
This question evaluates competencies in ML system design, similarity and representation modeling, feature and data engineering, scalable low-latency serving architectures, and privacy, fairness, and evaluation practices.
ML System DesignOnsite - GleanMediumMachine Learning EngineerDesign a Work-Together Distance Between Any Two Employees of a Company
Design a learned distance between any two employees of a company that reflects how much they actually work together. Tests ML framing without ground-truth labels, collaboration signals and their normalization, link-prediction training, evaluation, privacy and serving at company scale.
ML System DesignOnsite - GleanMediumMachine Learning EngineerFind Earliest Train Route
This question evaluates a candidate's ability to model and search time-dependent transit networks, combining temporal graph reasoning, state-space search, and path reconstruction for itinerary reporting.
Coding & AlgorithmsOnsite - GleanMediumMachine Learning EngineerImplement Rate-Limited Wikipedia Crawler
This question evaluates a candidate's ability to implement a rate-limited web crawler, testing skills in crawl frontier management, prioritization by page-title initial, deduplication, rate limiting, and robust error handling.
Coding & AlgorithmsOnsite - GoogleMediumMachine Learning EngineerCompare NLP tokenization and LLM recommendations
This question tests the conceptual understanding of tokenization strategies in NLP and the architectural trade-offs when integrating LLMs into recommendation systems.
Machine LearningOnsiteNLPLLMRecommendation-Systems - GoogleMediumMachine Learning EngineerCompute sum over consecutive-step subarrays
This problem tests your ability to identify contiguous subarrays that follow an arithmetic progression with a common difference of 1 (consecutive steps) and calculate the sum of elements within these subarrays efficiently.
Coding & AlgorithmsOnsiteArraysMathSliding Window - GoogleHardMachine Learning EngineerDescribe conflict and failure using STAR framework
This question evaluates interpersonal and leadership competencies such as conflict resolution, accountability, communication, and reflective learning, along with proficiency in structuring responses using the STAR behavioral framework.
Behavioral & LeadershipOnsite - GoogleHardMachine Learning EngineerDesign a real-time recommendation system
This question evaluates a candidate's competency in designing large-scale, low-latency real-time recommendation systems, covering feature engineering and feature-store consistency, candidate generation and ranking architectures, cold-start handling, latency versus accuracy trade-offs, monitoring, and operational scalability.
ML System DesignOnsite - GoogleMediumMachine Learning EngineerDesign a watch-next video recommender and derive its losses and optimizer update rules
Design the watch-next recommendations shown while a user watches a video, from objectives and candidate generation to ranking, serving and evaluation. The interview probes modeling depth: exact loss functions for retrieval and multi-task ranking, the gradient at the output logit, and the update rules of the optimizers used to train each model.
ML System DesignOnsite - GoogleHardMachine Learning EngineerDesign feedback-driven recommender
Design feedback-driven recommender 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 - GoogleHardMachine Learning EngineerDesign large-scale near-duplicate video detection
This question tests the ability to design an end-to-end pipeline for detecting near-duplicate videos at a massive scale, focusing on embedding generation, vector indexing, and trade-offs between precision and recall.
System DesignOnsiteComputer VisionEmbeddingsANN Search - GoogleHardMachine Learning EngineerExplain ML model fundamentals
Explain ML model fundamentals 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 - GoogleMediumMachine Learning EngineerGenerate values by weighted probabilities
This question evaluates understanding of probabilistic sampling, numerical stability, input validation, and algorithmic trade-offs for streaming weighted random draws, including generator determinism via seeding and methods such as CDF/binary-search versus alias for performance.
Statistics & MathOnsite - GoogleHardMachine Learning EngineerSolve several streaming, DAG, and DP tasks
Implement an ML pipeline optimization engine that calculates critical path latency and peak parallelism for a task DAG, and optimizes streaming checkpoint placement under sliding-window constraints using dynamic programming.
Coding & AlgorithmsOnsiteDAGDynamic ProgrammingSliding Window - HarveyMediumMachine Learning EngineerTraining LLM Agents for Long-Context, Long-Horizon Tasks: Environments and Rewards
An ML research discussion on training language-model agents for long-context, long-horizon tasks. It tests the training recipe from context extension through supervised trajectories to RL, the design of reproducible training environments, and reward design that resists reward hacking over long trajectories.
Machine LearningOnsite - HebbiaHardMachine Learning EngineerDesign a Multi-Agent Document Search and Visualization Product
This problem tests your ability to design a complex multi-agent orchestration system that integrates large-scale document retrieval with reliable citation tracking and a user-facing visualization interface.
ML System DesignOnsiteMulti-Agent SystemsRAGOrchestration - Hudson River TradingEasyMachine Learning EngineerCount People Reaching the Boundary
This question evaluates algorithmic simulation and event-driven modeling skills, specifically handling discrete versus continuous time, visibility constraints relative to a stationary watcher that changes facing direction, and boundary conditions for state transitions.
Coding & AlgorithmsOnsite - LiftoffHardMachine Learning EngineerProbability That the Sum of 1,000 Fair Die Rolls Exceeds a Threshold
A fair six-sided die is rolled 1,000 times; estimate the probability that the sum of the faces exceeds a threshold, including a threshold of 600. Tests the mean and variance of one roll, how they scale for a sum of independent rolls, the normal approximation with a continuity correction, and range checks.
Statistics & MathOnsite - LilaMediumMachine Learning EngineerDebug a Transformer bug in an unfamiliar repo
This question tests your ability to identify and rectify a logical error in a Transformer's multi-head self-attention mechanism, specifically focusing on tensor dimensionality, causality masking, and numerical stability.
Software Engineering FundamentalsOnsiteTransformersDeep LearningDebugging - LilaMediumMachine Learning EngineerDesign a DNA-sequence optimization loop
This problem tests your ability to design a closed-loop active learning system for protein or DNA sequence engineering, focusing on the iterative cycle of sequence generation, uncertainty estimation, and experimental validation.
ML System DesignOnsiteActive LearningGenerative ModelsBioinformatics
Figma
Glean
Google
Harvey
Hudson River Trading
Liftoff