ModelOp Center Glossary

Term	Definition
Abstraction	The art of replacing specific details (e.g. this model reads from an S3 bucket) with generic ones (this model reads data from a data handler, which may be S3 but may be something else entirely).
Asset	Any individual component that is used and required during a deployment in an engine. This includes models, schemas, streams, sensors, and import policies.
Attachments	A tarfile that contains all dependencies that the model being deployed needs to run properly.
Deployment	Making a model available for use by the business.
Engine	A docker container that executes model code that has been encapsulated into the ModelOp Center core abstractions: a model (the code), schemas, and streams.
Fleet	A list of active ModelOp Center containers that the user could interact with. This could include multiple ModelOp Center Engines and ModelOp Center Manage.
Jet	A Unix process that runs a model.
Job	A complete configuration of one or more interrelated ModelOp Center engines that each contain a model, schemas, and input/output stream(s).
Manifold	A component of an engine that manages the data flow between streams and the model.
Model Life Cycle (MLC)	The description of a model’s journey from creation, through testing, approvals, packaging, deployment and into maturity including monitoring, iteration and retirement.
Runner	ModelOp Center Engine will use different model-specific runners to execute depending on the language of the model deployed.
Sensor	A configurable function that captures specific metadata about the execution process of a model in production.
Schema	The definition of a Model’s expected data inputs or outputs expressed in a standard way. If the model is a function, the schemas define that function’s type signature.
Stream	A file that contains all configuration information necessary to transport data from one place to another. The transport could be from a data source to the engine or from the engine to an application. There is at least one input stream and one output stream required for model execution.
Training	Tuning model parameters to optimize performance on a particular dataset. Training results in an input set of model parameters as well as the corresponding output model artifacts (coefficients, weights, etc.)