Agents

Autonomous systems that perform tasks and interact with users or the environment.

Chain of Thought (CoT) Prompting

Prompting with step-by-step instructions to show "how".

Embedding

A way of representing words or data as numerical vectors.

Few-shot Learning

Training AI to perform tasks using only a few examples.

Hallucination

When a model generates false or incorrect information.

HuggingFace

Website with tools and models for machine learning.

MCP

LLM to API translation server.

Mixture of Experts (MOE)

A machine learning model that routes tasks to specialized sub-models.

Neural Network

Mathematical system that identifies and analyzes statistical patterns.

NIM (Nvidia)

NVIDIA Networking Infrastructure Management.

PRD

A Product Requirements Document.

Prompt Engineering

Crafting inputs to guide AI models toward desired outputs.

RAG

Adding new information to supplement model knowledge.

Reinforcement Learning

Machine learning where an agent learns through trial and error.

Reward Modeling

Training AI by defining rewards for desirable behaviors.

Supervised Learning

Machine learning where an algorithm learns from labeled datasets.

Temperature

Controls the randomness of text generation.

Token

A basic unit of text in AI models, such as a word or subwords.

Transformers

Models that grasp context and the language’s long-term associations

Unsupervised Learning

Machine learning where AI finds patterns without labeled data.

Vectorize

Process for converting data into numerical vectors (Embedding).