Application
This is the part of the AI stack that users actually interact with, often taking the form of a chatbot, a dashboard, or a mobile app
User Interface (UI)
This process determines the order in which different AI tasks and tools are executed in a workflow
Task Sequencing
Text, images, audio, video, and numbers are all examples of this core resource used by AI systems
Data
This is the process where an AI model learns patterns from data so it can make prediction
Training
These machines provide the computing power needed to run AI models and applications
Often called the "bridge" between layers, this acronym refers to the set of rules that allow the Application Layer to send a request to a Large Language
API (application programming interface)
When one AI system automatically decides which model, API, or tool to call next, it is doing this
Before data is fed into an AI model, it often goes through this process to be cleaned, organized, and prepared
Data Processing
This type of dataset is used to evaluate how well a trained model performs on unseen data
Test Set
These processors are commonly used in AI because they can handle many operations in parallel more efficiently than standard CPUs
GPUs
To make AI useful for a specific company, developers add this at the application layer to ensure the AI follows company policies, legal regulations, and safety standards
Business Logic
This step in an AI workflow combines results from different tools or model calls into one final output.
Aggregation
When the training data is unfair, incomplete, or unbalanced, AI systems can develop this problem in their outputs
This metric measures the proportion of correct predictions made by a model out of all predictions
Accuracy
When companies run AI systems on remote servers instead of their own physical machines, they are using this
Cloud Computing
This term describes the invisible instructions embedded in the application layer that guide the AI's persona, tone, and constraints, often hidden from the end-user's view
System Prompt (system message)
If an AI assistant retrieves data, sends it to a model, and then triggers another tool based on the result, this overall coordination is an example of this
Pipeline Orchestration
This kind of data is organized into rows and columns, making it easier to store, search, and analyze than raw text or images
Structured Data
When a model performs extremely well on training data but poorly on new data, it is experiencing this issue
Overfitting
This refers to the ability of a system to handle increasing workloads by adding more resources or machine
Scalability
At the highest level of the application layer, these autonomous systems can not only generate text but also use "tools" to execute multi-step tasks like booking a flight or updating a CRM without human intervention
AI Agent
This orchestration concept focuses on connecting multiple tools or services so they act together as one larger automated process
System Integration
This principle refers to knowing where a dataset came from, how it was changed, and how it moved through a system over time
Data Lineage
This technique improves model performance by combining predictions from multiple models instead of relying on a single one
Ensemble Learning
This container management platform is widely used to deploy, run, and scale AI applications across multiple servers
Kubernetes