You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
description: In this section, we provide a high-level explanation of the parts of the product and explain how these parts work together. We also show how the product would work when solving the scenario outlined in the introduction section.
description: In this section, we provide a high-level explanation of the parts of the product and explain how these parts work together. We also show how the product would work when solving the scenario outlined in the introduction section.
description: In this section, we create interactive questions that validate if the learner has understood the learning objectives
7
-
ms.date: 11/28/2023
8
-
author: PatAltimore
9
-
ms.author: leestott
10
-
ms.topic: unit
11
-
ms.custom: team=nextgen
12
-
durationInMinutes: 5# dummy number at present
13
-
content: |
14
-
quiz:
15
-
questions:
16
-
- content: "If you want to deploy MLOps, what is the correct sequence of deployment?"
17
-
choices:
18
-
- content: "Build and Train; Package and Deploy; Monitor and Retrain"
19
-
isCorrect: true
20
-
explanation: "Correct. You first create the model, then you can package and deploy the model. The model is monitored and retrained as needed."
21
-
- content: "Monitor and Retrain; Package and Deploy; Build and Train"
22
-
isCorrect: false
23
-
explanation: "Incorrect. Build and Train is the first step in the deployment of MLOps."
24
-
- content: "Package and Deploy; Monitor and Retrain; Build and Train"
25
-
isCorrect: false
26
-
explanation: "Incorrect. You can't package and deploy a model without building it first."
27
-
- content: "MLOps brings the philosophy of DevOps to the Data science community. In the absence of MLOPs, how would the work of a data scientist be impacted?"
28
-
choices:
29
-
- content: "A data scientist won't be able to conduct experiments."
30
-
isCorrect: false
31
-
explanation: "Incorrect. A data scientist can conduct experiments without MLOps."
32
-
- content: "A data scientist won't be able to test models."
33
-
isCorrect: false
34
-
explanation: "Incorrect. A data scientist can test models without MLOps."
35
-
- content: "A data scientist wouldn't be able to collaborate with developers efficiently."
36
-
isCorrect: true
37
-
explanation: "Correct. A data scientist wouldn't be able to collaborate effectively with developers in the absence of MLOps because they would be each following different workflow."
38
-
- content: "A company has decided to implement MLOps on Edge devices. Which step would they need to implement before deploying MLOPs on Edge devices?"
39
-
choices:
40
-
- content: "The company should deploy IoT Edge and capture data from sensors."
41
-
isCorrect: true
42
-
explanation: "Correct. Without IoT Edge being operational, you can't acquire data to build models."
43
-
- content: "The company should be able to visualize data from sensors."
44
-
isCorrect: false
45
-
explanation: "Incorrect. The ability to visualize data isn't a prerequisite for MLOps on edge devices."
46
-
- content: "Developers should follow CI/CD processes before MLOps is deployed."
47
-
isCorrect: false
48
-
explanation: "Incorrect. The ability to use CI/CD isn't a prerequisite for MLOps on edge devices."
description: In this section, we create interactive questions that validate if the learner has understood the learning objectives
7
+
ms.date: 05/12/2025
8
+
author: PatAltimore
9
+
ms.author: leestott
10
+
ms.topic: unit
11
+
ms.custom: team=nextgen
12
+
durationInMinutes: 5
13
+
content: |
14
+
quiz:
15
+
questions:
16
+
- content: "If you want to deploy MLOps, what is the correct sequence of deployment?"
17
+
choices:
18
+
- content: "Build and Train; Package and Deploy; Monitor and Retrain"
19
+
isCorrect: true
20
+
explanation: "Correct. You first create the model, then you can package and deploy the model. The model is monitored and retrained as needed."
21
+
- content: "Monitor and Retrain; Package and Deploy; Build and Train"
22
+
isCorrect: false
23
+
explanation: "Incorrect. Build and Train is the first step in the deployment of MLOps."
24
+
- content: "Package and Deploy; Monitor and Retrain; Build and Train"
25
+
isCorrect: false
26
+
explanation: "Incorrect. You can't package and deploy a model without building it first."
27
+
- content: "MLOps brings the philosophy of DevOps to the Data science community. In the absence of MLOPs, how would the work of a data scientist be impacted?"
28
+
choices:
29
+
- content: "A data scientist won't be able to conduct experiments."
30
+
isCorrect: false
31
+
explanation: "Incorrect. A data scientist can conduct experiments without MLOps."
32
+
- content: "A data scientist won't be able to test models."
33
+
isCorrect: false
34
+
explanation: "Incorrect. A data scientist can test models without MLOps."
35
+
- content: "A data scientist wouldn't be able to collaborate with developers efficiently."
36
+
isCorrect: true
37
+
explanation: "Correct. A data scientist wouldn't be able to collaborate effectively with developers in the absence of MLOps because they would be each following different workflow."
38
+
- content: "A company has decided to implement MLOps on Edge devices. Which step would they need to implement before deploying MLOPs on Edge devices?"
39
+
choices:
40
+
- content: "The company should deploy IoT Edge and capture data from sensors."
41
+
isCorrect: true
42
+
explanation: "Correct. Without IoT Edge being operational, you can't acquire data to build models."
43
+
- content: "The company should be able to visualize data from sensors."
44
+
isCorrect: false
45
+
explanation: "Incorrect. The ability to visualize data isn't a prerequisite for MLOps on edge devices."
46
+
- content: "Developers should follow CI/CD processes before MLOps is deployed."
47
+
isCorrect: false
48
+
explanation: "Incorrect. The ability to use CI/CD isn't a prerequisite for MLOps on edge devices."
MLOps (DevOps for machine learning) enables data science, and IT teams to collaborate and increase the pace of model development and deployment via monitoring, validation, and governance of machine learning models. MLOps brings the philosophy of DevOps to Machine Learning by automating the end-to-end workflows. MLOps is a part of Azure Machine Learning and can be deployed to IoT Edge devices.
1
+
MLOps (DevOps for machine learning) lets data science and IT teams collaborate to speed up model development and deployment through monitoring, validation, and governance of machine learning models. MLOps applies the philosophy of DevOps to machine learning by automating end-to-end workflows. MLOps is a part of Azure Machine Learning and can be deployed to IoT Edge devices.
2
2
3
-
The end-to-end Machine Learning life cycle is seen below:
3
+
The end-to-end machine learning lifecycle is shown in the following diagram:
4
4
5
5
6
6
7
-
1. Create models with reusable Machine Learning pipelines using the Azure Machine Learning extension for Azure DevOps. Store your code in GitHub, so it automatically integrates into your MLOps pipeline.
7
+
1. Create models with reusable machine learning pipelines using the Azure Machine Learning extension for Azure DevOps. Store your code in GitHub so it integrates automatically into your MLOps pipeline.
8
8
9
9
2. Automate your MLOps rollout using Azure DevOps + Azure Machine Learning for version models with rich metadata and event management.
10
10
11
-
3. Automatically create an audit trail for all artifacts in your MLOPs pipeline ensure asset integrity and meet regulatory requirements.
11
+
3. Automatically create an audit trail for all artifacts in your MLOps pipeline to ensure asset integrity and meet regulatory requirements.
12
12
13
13
4. Deploy and monitor performance so you can release models with confidence and know when to retrain.
14
14
@@ -18,16 +18,16 @@ The deployment of machine learning models in production presents one of the most
18
18
19
19
- IoT machine learning models are rapidly changing, hence they degrade faster (with respect to data drift of the current data). Therefore, they need more frequent and automatic retraining.
20
20
21
-
- IoT machine learning models need to be deployed on different kind of target platforms, and you need to leverage the capabilities of these platforms in terms of performance, security, so on.
21
+
- IoT machine learning models need to be deployed on different kinds of target platforms, and you need to use the capabilities of these platforms for performance, security, and so on.
22
22
23
-
- IoT Edge solutions may need to run offline – hence you need to allow for offline working with the frequency of refresh for the models.
23
+
- IoT Edge solutions might need to run offline, so you need to allow for offline operation with the frequency of model refresh.
24
24
25
25
## Scenario
26
26
27
-
Suppose you work in the Oil and Gas industry, and you're responsible for maintaining thousands of oil and gas pumps operating in remote/offshore locations. Even if these locations are remote, your team must rapidly identify and fix faults in the field. Hence, you want to build and deploy a predictive maintenance system for the pumps. Before you create a predictive maintenance model, you must first implement a system for remote monitoring of connected devices to ingest sensor data in a stream, process that data, and store it in a database. Currently, your team has managed to achieve this goal with IoT Edge. You can capture data from multiple sensors in the field connected to the pumps. Currently, anomalies are detected visually from a dashboard. Hence, your engineers want to deploy a new predictive maintenance system that can detect anomalies from up-to-date machine learning models. The models generated by the new predictive maintenance system should reflect the current state of the data dynamically by accounting for data drift. Hence, the system should allow for frequent and automatic retraining of models to indicate the status. You also need to deploy the trained models on a variety of pumps from a range of manufacturers. The system should use the unique characteristics of the software on the pumps: both for performance and security. Finally, the models deployed should be able to run offline on the Edge device if needed. The company can avail substantial savings on maintenance and production costs and increase workplace safety and their environmental impact by achieving these objectives.
27
+
Suppose you work in the oil and gas industry, and you're responsible for maintaining thousands of pumps operating in remote or offshore locations. Even if these locations are remote, your team must rapidly identify and fix faults in the field. So, you want to build and deploy a predictive maintenance system for the pumps. Before you create a predictive maintenance model, you must first implement a system for remote monitoring of connected devices to ingest sensor data in a stream, process that data, and store it in a database. Your team achieves this goal with IoT Edge. You can capture data from multiple sensors in the field connected to the pumps. Anomalies are detected visually from a dashboard. So, your engineers want to deploy a new predictive maintenance system that detects anomalies from up-to-date machine learning models. The models generated by the new predictive maintenance system reflect the current state of the data dynamically by accounting for data drift. Hence, the system should allow for frequent and automatic retraining of models to indicate the status. You also need to deploy the trained models on various pumps from a range of manufacturers. The system should use the unique characteristics of the software on the pumps: both for performance and security. Finally, the models deployed should be able to run offline on the Edge device if needed. The company can achieve substantial savings on maintenance and production costs, increase workplace safety, and reduce its environmental impact by achieving these objectives.
28
28
29
29
30
30
31
-
This module covers **source control, reproducible training pipeline, model storage, and versioning**, **model packaging**, **model validation**, **deployment,****monitoring models in production, and retraining of models** in context for IoT Edge devices.
31
+
This module covers **source control, reproducible training pipelines, model storage and versioning**, **model packaging**, **model validation**, **deployment**,**monitoring models in production**, and **retraining of models** in the context of IoT Edge devices.
32
32
33
-
In this module, you'll examine the significance of MLOps for IoT Edge in the context of the above scenario for the development and deployment of machine learning models.
33
+
In this module, you examine the significance of MLOps for IoT Edge in the context of the previous scenario for developing and deploying machine learning models.
0 commit comments