# MLOps

> The tools and habits that keep an AI model working. They start once people use it.

**Field:** Artificial Intelligence  
**Category:** MLOps & Deployment  
**Also known as:** Machine Learning Operations, ML Ops

## Think of it like

Think of it like running a bake shop, not baking one cake. You need the same good bread each day.

## Definition

MLOps is the set of habits and tools for running machine learning models in real use, not just training them once. It takes ideas from DevOps, such as tests, version control and doing jobs by code. Then it adds what models need. That means tracking data, training again, storing models, serving them and watching them once they are live. The goal is a model you can rebuild, check and update with no risk. Most models fail after launch for lack of this, not for lack of a good first try.

## Example

Lucy's team built a good churn model, but it took a week to retrain by hand. After setting up MLOps, a new version trains, tests and ships each Monday on its own.

## Why it matters

A model that works once on a laptop is a demo. MLOps is what turns it into a product people can trust.

## Common mistake

Treating the first trained model as the finish line. The world keeps changing, so the model needs care for as long as it is used.

## Three principles

- **Like DevOps.** For models.
- **Repeatable.** Rebuild it any time.
- **Watch it live.** Then improve it.

## Related terms

- [DevOps](https://learn.benibix.com/term/development/devops)
- [ML Pipeline](https://learn.benibix.com/term/ai/ml-pipeline)
- [Model Monitoring](https://learn.benibix.com/term/ai/model-monitoring)
- [Machine Learning](https://learn.benibix.com/term/ai/machine-learning)
- [Amazon SageMaker AI](https://learn.benibix.com/term/ai/amazon-sagemaker-ai)
- [Large Language Model Operations (LLMOps)](https://learn.benibix.com/term/ai/llmops)
- [Automated Machine Learning (AutoML)](https://learn.benibix.com/term/ai/automl)

## Told apart from

- [MLOps vs DevOps](https://learn.benibix.com/compare/mlops-vs-devops): People often think MLOps is just DevOps used on a model. It takes a lot from DevOps, such as tests, version control and doing jobs by code. But a model also depends on its data. So MLOps has to track data versions, train again, and watch a live model for drift. A normal app does not get worse by itself when the world changes, but a model does.
- [MLOps vs LLMOps](https://learn.benibix.com/compare/mlops-vs-llmops): People think LLMOps is a new field that replaces MLOps, and it is not. It is the same habit for apps built on a language model. The loop of test, ship and watch stays. What changes is the part you edit. In MLOps you retrain the model. In LLMOps you rarely do, so you version the prompt, test it with saved questions and trace each call.

![A loop with four stops: data, train, deploy and monitor. An arrow runs from monitor back to data, so the model is improved again and again.](https://learn.benibix.com/diagrams/mo-mlops.svg)

---

Source: https://learn.benibix.com/term/ai/mlops
Author: Behnia Zarei (a developer in Vienna, Austria, who builds and teaches data, AI and business intelligence at BENIBIX)
Publisher: BENIBIX Learn (https://learn.benibix.com/)
Last updated: 2026-11-17
Licence: CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). Quote freely with attribution.
