AI and LLM engineering practice
Every AI and LLM engineering exercise, built from a topic you can open beside it.
75 exercises runs offline
All tracks Python 139 JavaScript 184 TypeScript 76 Go 106 Rust 75 Java 81 Kotlin 66 C++ 92 C# 88 Swift 68 PHP 69 Frontend 82 Backend 139 Architecture and system design 25 DevOps and cloud 15 Data and databases 5 Data science 31 AI and LLM engineering 75 Security 26 Game development 26 Coding in the AI era 101 CS foundations 164
level
llm Quiz new A model returns a well-formed request to refund an order. Which component should make the… Quiz llm Quiz new Why does scaled dot-product attention divide query-key scores by the square root of the k… Quiz llm Fill in new Name the training procedure that applies the chain rule backward through a computation gr… Fill in llm Quiz new A classifier prompt repeatedly confuses two labels. Which change is the most targeted few… Quiz llm Fill in new Name the mechanism added to token representations so a Transformer can distinguish sequen… Fill in llm Fill in new Name the failure mode in which untrusted input tries to override application instructions… Fill in llm Quiz new What does a ReLU activation output for inputs -2 and 3? Quiz llm Fill in new Name the two-word term for the explicit definition of an NLP input, output, labels, metri… Fill in llm Fill in new Name the two-word unit used in Claude message content arrays for text, tool use, and tool… Fill in llm Fill in new Name the durable request identifier used to prevent a retry from applying the same side e… Fill in llm Quiz new What does plain JSON mode guarantee? Quiz llm Fill in new Name the schema language commonly used to declare structured-output fields, types, requir… Fill in llm Quiz new Which statement correctly separates MCP host, client, and server responsibilities? Quiz llm Quiz new How should an OpenAI Responses API client extract generated text reliably? Quiz llm Quiz new A model emits a function call with schema-valid arguments. What should happen next? Quiz llm Fill in new Which Claude content-block type returns a client tool outcome and pairs it with the origi… Fill in llm Fill in new Name the ingestion step that splits long source documents into retrievable passages Fill in llm Fill in new Name the metric that compares vector direction by dividing the dot product by both vector… Fill in llm Quiz new What does a high similarity score between two text embeddings establish? Quiz llm Quiz new What is the usual effect of increasing the HNSW query-time search-width parameter ef? Quiz llm Fill in new Write the ordered three-part form used to represent a fact in a knowledge graph Fill in llm Quiz new A product manual changes every week. Why can RAG be preferable to fine-tuning for answeri… Quiz llm Fill in new Name the retrieval metric that asks whether relevant evidence appears among the first k r… Fill in llm Quiz new What does wrapping a function in a LangChain RunnableLambda guarantee? Quiz llm Fill in new Name the cyclic control structure that sends tool results or screenshots back to a model … Fill in llm Fill in new Name the persistence mechanism that saves graph state so a long-running agent can resume … Fill in llm Fill in new Name the control pattern that pauses a computer-use agent for a person to approve a high-… Fill in llm Quiz new Which description matches LangGraph execution? Quiz llm Quiz new What is the defining purpose of multimodal fusion? Quiz llm Quiz new Which execution pattern characterizes a ReAct-style agent? Quiz llm Fill in new Name the VLM component that converts pixels or image patches into visual feature represen… Fill in llm Quiz new A VLM reads an invoice total from a low-resolution screenshot before payment. What is the… Quiz llm Quiz new A team wants to know whether a prompt revision improved its application. Which evaluation… Quiz llm Fill in new Name the failure mode in which evaluation examples or their answers influence training, t… Fill in llm Quiz new Which Hugging Face loading practice gives the strongest reproducibility boundary? Quiz llm Quiz new Which architecture best limits damage if a jailbreak or prompt injection succeeds? Quiz llm Fill in new Name the parameter-efficient fine-tuning method that learns low-rank update matrices whil… Fill in llm Quiz new Which observability record is most useful and safest for debugging a slow RAG answer? Quiz llm Quiz new A quantized local model file fits in GPU memory. What must capacity planning still includ… Quiz llm Fill in new Name the observability unit that represents one complete application request as related s… Fill in llm Quiz new What must a multi-turn application send on the next stateless Messages API request? Build reliable Claude Messages API requests, content-block parsers, streams, and tool loops with safe retry boundaries. llm Predict the output new What does this content-block parser print? Build reliable Claude Messages API requests, content-block parsers, streams, and tool loops with safe retry boundaries. llm Quiz new A charge-card tool times out after the remote service may have accepted the charge. What … Build reliable Claude Messages API requests, content-block parsers, streams, and tool loops with safe retry boundaries. llm Review AI code new Review generated refund handler Build reliable Claude Messages API requests, content-block parsers, streams, and tool loops with safe retry boundaries. llm Spot the bug new Spot the bug in tool continuation Build reliable Claude Messages API requests, content-block parsers, streams, and tool loops with safe retry boundaries. llm Quiz new Why should a team avoid changing the prompt and model in the same evaluation run? Build your first LLM feature with an explicit input contract, output validation, an evaluation set, and a safe model boundary. llm Predict the output new What does this validation fragment print? Build your first LLM feature with an explicit input contract, output validation, an evaluation set, and a safe model boundary. llm Review AI code new Review generated ticket handler Build your first LLM feature with an explicit input contract, output validation, an evaluation set, and a safe model boundary. llm Spot the bug new Spot the bug in the response parser Build your first LLM feature with an explicit input contract, output validation, an evaluation set, and a safe model boundary. llm Quiz new A model response is valid JSON and matches its response schema. What must happen before i… Build your first LLM feature with an explicit input contract, output validation, an evaluation set, and a safe model boundary. llm Quiz new Which loader expresses that a checkpoint must provide a sequence-classification head? Use the Hub to manage model artifacts and Transformers Pipeline and AutoClass for reproducible, secure local inference. llm Predict the output new What does this revision-consistency check print? Use the Hub to manage model artifacts and Transformers Pipeline and AutoClass for reproducible, secure local inference. llm Quiz new When should a service enable trust_remote_code for a Hub model? Use the Hub to manage model artifacts and Transformers Pipeline and AutoClass for reproducible, secure local inference. llm Review AI code new Review generated private classifier Use the Hub to manage model artifacts and Transformers Pipeline and AutoClass for reproducible, secure local inference. llm Spot the bug new Spot the bug in the model loader Use the Hub to manage model artifacts and Transformers Pipeline and AutoClass for reproducible, secure local inference. llm Quiz new Two RunnableBranch predicates both match an input. Which branch runs? A practical guide to LangChain 1.x runnables, model pipelines, tool-calling agents, state boundaries, and migration traps. llm Quiz new When is create_agent a better fit than a fixed runnable sequence? A practical guide to LangChain 1.x runnables, model pipelines, tool-calling agents, state boundaries, and migration traps. llm Predict the output new What does this runnable sequence print? A practical guide to LangChain 1.x runnables, model pipelines, tool-calling agents, state boundaries, and migration traps. llm Review AI code new Review generated refund agent A practical guide to LangChain 1.x runnables, model pipelines, tool-calling agents, state boundaries, and migration traps. llm Spot the bug new Spot the bug in async chain invocation A practical guide to LangChain 1.x runnables, model pipelines, tool-calling agents, state boundaries, and migration traps. llm Quiz new What must you verify before claiming that a local LLM deployment is offline? Select, run, and evaluate LLMs on your own hardware while handling quantization, context memory, streaming, and service boundaries. llm Predict the output new What does this program print? Select, run, and evaluate LLMs on your own hardware while handling quantization, context memory, streaming, and service boundaries. llm Quiz new Which statement is safe to use when planning memory for a quantized local model? Select, run, and evaluate LLMs on your own hardware while handling quantization, context memory, streaming, and service boundaries. llm Review AI code new Review generated private summarizer Select, run, and evaluate LLMs on your own hardware while handling quantization, context memory, streaming, and service boundaries. llm Spot the bug new Spot the bug in the stream collector Select, run, and evaluate LLMs on your own hardware while handling quantization, context memory, streaming, and service boundaries. llm Quiz new A product request says only “understand support messages.” What should the NLP team defin… Turn language into testable classification, labeling, and retrieval tasks without losing semantics, offsets, or evaluation integrity. llm Quiz new A browser receives { start: 12, end: 18 } from a Python entity service. What is missing f… Turn language into testable classification, labeling, and retrieval tasks without losing semantics, offsets, or evaluation integrity. llm Predict the output new What does this normalization code print? Turn language into testable classification, labeling, and retrieval tasks without losing semantics, offsets, or evaluation integrity. llm Review AI code new Review generated ticket classifier Turn language into testable classification, labeling, and retrieval tasks without losing semantics, offsets, or evaluation integrity. llm Spot the bug new Spot the bug in the feature pipeline Turn language into testable classification, labeling, and retrieval tasks without losing semantics, offsets, or evaluation integrity. llm Quiz new A request returns HTTP 429 with error.code organization_spend_limit_exceeded. What is the… Build with the Responses API, parse typed output, manage conversations and streams, and handle keys, status, and retry boundaries correctly. llm Quiz new A second Responses API call uses previous_response_id. What should it do with trusted ins… Build with the Responses API, parse typed output, manage conversations and streams, and handle keys, status, and retry boundaries correctly. llm Predict the output new What does this response-item parser print? Build with the Responses API, parse typed output, manage conversations and streams, and handle keys, status, and retry boundaries correctly. llm Review AI code new Review generated support summary publisher Build with the Responses API, parse typed output, manage conversations and streams, and handle keys, status, and retry boundaries correctly. llm Spot the bug new Spot the bug in the Responses API reader Build with the Responses API, parse typed output, manage conversations and streams, and handle keys, status, and retry boundaries correctly.
No practice items match these filters.