THE COMPLETE CURRICULUM

A clearer picture.
One concept at a time.

Follow the six levels in order, or jump to the question on your mind.

LEVEL 01 · 10 LESSONS

Understand AI

Build the mental models behind the answers.

01TokensTokens are the small units a language model processes.3 min○02Context windowsThe context window is the model’s capacity for information within one request.3 min○03TemperatureTemperature is a generation setting that changes how strongly a model favors likely next tokens.3 min○04EmbeddingsAn embedding is a list of numbers representing features of an input, such as text.3 min○05Embedding models vs LLMsAn embedding model produces representations useful for comparison.3 min○06Deterministic vs probabilistic systemsA deterministic rule produces the same result from the same inputs under the same conditions.3 min○07HallucinationsA hallucination is an output that is fabricated, unsupported, or inconsistent with the relevant evidence.3 min○08Model families and sizesA model family is a related group of models.3 min○09Open vs closed modelsModels vary in how much is available to inspect, modify, and run.3 min○10Inference vs trainingTraining changes a model’s learned parameters using data.3 min○

LEVEL 02 · 10 LESSONS

Give AI knowledge

Connect answers to the information that matters.

11GroundingGrounding means anchoring an answer in relevant evidence supplied to the system, such as a policy, a database result, or an observed tool response..3 min○12Retrieval-augmented generation (RAG)RAG combines finding relevant information with generating an answer from it.3 min○13ChunkingChunking splits larger source material into smaller units for indexing and retrieval.3 min○14Semantic chunkingSemantic chunking chooses boundaries using changes in meaning or topic, rather than only a fixed character or token count..3 min○15MetadataMetadata is descriptive information attached to content: its source, owner, date, region, document type, section, or access classification..3 min○16Vector databasesA vector database stores numerical representations and supports searching for nearby vectors.3 min○17Similarity searchSimilarity search ranks items by how close their representations are to a query.3 min○18RerankingReranking is a second pass that reorders retrieved candidates using a more focused assessment of their relevance to the question..3 min○19Parent-child retrievalParent-child retrieval searches smaller passages but returns a larger containing section when more context is needed..3 min○20Handling cross-referencesA cross-reference points from one part of a source to another, such as ‘subject to section 8.’ Retrieval must preserve or follow that relationship when the referenced material affects the answer..3 min○

LEVEL 03 · 8 LESSONS

Shape the behavior

Turn a broad capability into a useful response.

21System promptsA system prompt is an instruction supplied by the application to define the model’s role, priorities, and expected behavior.3 min○22Few-shot promptingFew-shot prompting supplies a small set of examples showing the desired input-to-output pattern within the request..3 min○23Structured outputStructured output returns information in a defined format, such as a record with fixed fields and allowed categories, rather than only free-form prose..3 min○24Fine-tuningFine-tuning is additional training that adjusts a model using a selected dataset to improve a target behavior or task..3 min○25Confidence thresholdsA confidence threshold determines when a system accepts a result and when it asks for clarification, review, or a different method.3 min○26FallbacksA fallback is a planned alternative when the preferred AI path cannot complete a task reliably or safely..3 min○27Prompt templatesA prompt template is a reusable structure that combines stable instructions with task-specific inputs..3 min○28Model routingModel routing selects a model or processing path for a task instead of sending every request to the same model..3 min○

LEVEL 04 · 8 LESSONS

Let AI do work

Understand tools, workflows, and bounded autonomy.

29Tool callingTool calling lets a model request a defined operation from another system, such as looking up an order or preparing a calendar event..3 min○30AI agentsAn AI agent is a system that uses a model to choose and perform steps toward a goal, often using tools and feedback.3 min○31Agent loopsAn agent loop repeats a cycle of choosing a step, taking an allowed action, observing the result, and deciding what to do next..3 min○32Workflows vs agentsA workflow follows a predefined process.3 min○33Human-in-the-loopHuman-in-the-loop means a person participates at a defined point in an AI-assisted process to review, correct, approve, or resolve uncertainty..3 min○34MemoryApplication memory is information retained for use beyond the current interaction, such as preferences, summaries, or selected prior events..3 min○35StateState is the explicit record of where a task currently stands: inputs collected, steps completed, approvals received, actions taken, and work still pending..3 min○36OrchestrationOrchestration coordinates the models, tools, people, and services that make up a larger process..3 min○

LEVEL 05 · 9 LESSONS

Make it reliable

Evaluate quality and keep people in control.

37EvalsEvals are systematic tests of how well an AI system performs its intended task.3 min○38GuardrailsGuardrails are checks and controls that limit unacceptable inputs, outputs, or actions around an AI system..3 min○39Prompt injectionPrompt injection happens when untrusted content tries to redirect a model’s behavior by presenting itself as an instruction..3 min○40PermissionsPermissions define which identities may access particular data or perform particular operations.3 min○41Approval gatesAn approval gate pauses a proposed operation until an authorized person approves the specific action under the required conditions..3 min○42ObservabilityObservability is the ability to understand a system’s behavior from signals such as logs, metrics, traces, and user feedback..3 min○43TracingA trace connects the steps of a single task so you can follow its path through retrieval, model calls, tools, validations, and approvals..3 min○44Cost and latencyCost is the resource or monetary expense of a task.3 min○45Reliability vs autonomyAutonomy is how much a system can decide and do without intervention.3 min○

LEVEL 06 · 10 LESSONS

Navigate the ecosystem

Know what the pieces do, beyond the product names.

46Hugging FaceHugging Face provides a platform and tools for working with machine-learning models and datasets.3 min○47OllamaOllama is software for using supported models through local applications and APIs.3 min○48NVIDIA CosmosNVIDIA Cosmos is a platform and family of world-model capabilities aimed at physical AI: systems that interpret, generate, or reason about physical environments..3 min○49Local modelsA local model runs on hardware you operate, such as a laptop, workstation, or private server, rather than sending inference requests to an external model service..3 min○50Model servingModel serving is the operational layer that accepts requests, runs a model, and returns results to applications..3 min○51GPUs and inferenceGPUs are processors designed for large amounts of parallel computation.3 min○52APIsAn application programming interface, or API, is a defined way for software systems to request information or operations from each other..3 min○53Model Context Protocol (MCP)MCP is a protocol for connecting AI applications with external capabilities and context through a common interface..3 min○54AI gatewaysAn AI gateway is an intermediary through which applications access model services.3 min○55Agent frameworksAgent frameworks provide reusable building blocks for model calls, tools, task state, orchestration, and related agent behaviors.3 min○