The language of production AI, explained plainly.

Agentic RAG

RAG in which an AI agent decides how to search, for example rewriting the query, choosing sources or searching again, instead of running a single fixed retrieval step.

AI agent

An AI system that pursues a goal by deciding which actions or tools to use, step by step, rather than producing a single answer.

Chunking

Splitting documents into smaller passages for indexing, so retrieval returns focused, self-contained context.

Embeddings

Numeric vectors that represent the meaning of text, so that similar meanings end up close together and can be found with vector search.

Fine-tuning

Further training a pre-trained model on your own examples to change its behavior, style or format.

GraphRAG

A RAG approach that organizes knowledge as a graph of entities and relationships, so answers can follow connections across documents.

Guardrails

Checks around a language model that keep inputs and outputs within safe, allowed and useful bounds.

Hallucination

When a language model states something false or unsupported with confidence.

Human-in-the-loopHITL

A design in which an AI system pauses for a person to approve, edit or reject an action before it takes effect.

Hybrid search

Search that combines keyword matching (such as BM25) with semantic vector search, so results catch both exact terms and meaning.

Intelligent document processingIDP

Using OCR, machine learning and language models to turn documents such as scans, PDFs and forms into structured, usable data.

LangGraph

A framework for building stateful AI agents as graphs of steps, with checkpointed state and support for pausing for human input.

LLM evaluationRAGAS

Measuring the quality of an LLM or RAG system on a fixed set of real questions, with metrics such as faithfulness, relevance and retrieval hit-rate.

LoRA / QLoRA

Parameter-efficient fine-tuning methods that train small adapter matrices instead of the whole model, cutting compute and memory needs.

Model Context ProtocolMCP

An open standard for connecting AI applications to tools and data through reusable servers, instead of custom integrations per app.

Multi-agent system

Several specialized AI agents that divide a task between them, such as one extracting data, one summarizing and one structuring the output.

PII anonymization

Detecting personally identifiable information, such as names, addresses and ID numbers, and masking or replacing it before data is stored, shared or sent to a model.

Prompt caching

Reusing the processed form of a long, unchanging prompt prefix across requests, cutting input-token cost and latency.

Re-ranking

A second retrieval pass in which a more precise model re-orders the top search results before they reach the language model.

Retrieval-Augmented GenerationRAG

A technique that lets a language model answer from your own documents by retrieving relevant passages first and generating the answer from them.

Structured outputs

Constraining a language model to return data in a defined schema, such as JSON with specific fields, instead of free text.

Vector database

A database designed to store embeddings and quickly find the vectors most similar to a query.