Azure MLOps - DevOps for Machine Learning (Part 10| Last Part) | Testing End to End MLOps Pipelines

This Azure MLOps series contains 10 parts that show you how to apply DevOps for Machine Learning (MLOps) on Microsoft Azure.

If you have not watched the first video yet, I do recommend you start watching from Part 1:

The GitHub repo with codes I used for this series:

🔔 Subscribe for more cloud computing, data, and AI analytics videos
by clicking on subscribe button so you don't miss anything.

✅ You can contact me at:

LinkedIn:
Email:

#AzureML #AzureMachineLearning #MLOps #AzureMLOps #AzureDevOps #DevOpsForMachineLearning #MAchineLearningOperation #Azure #AI #DevOps

[adeebhakim]

What an excellent tutorial video on DevOps! Thanks MG.

[dharmendrasingh]

The best tutorials which covers DevOps for Machine Learning. Thank you!

[tolulopeolorunfemi]

Thanks for sharing this knowledge. It made me understand a lot of things in MLOPs

[myself]

🎯 Key Takeaways for quick navigation:

00:00 🎉 *Introduction to Testing MLOps Solution*
- Introduces the final video in the MLOps series, focusing on testing the MLOps solution.
- Sets the context for testing the automated MLOps lifecycle.
03:39 🔄 *Triggering CI Pipeline with Code Change*
- Demonstrates how a data scientist makes a change in the repository using VS Code.
- Simulates a scenario where a new development triggers the CI pipeline.
05:05 🔄 *Automated CI/CD Process*
- Observes the automatic progression of the CI pipeline based on the code change.
- Highlights the steps of the CI pipeline, including creating ML workspace, cluster, training, and model registration.
06:29 🔄 *CI/CD Deployment to Staging and Production*
- Shows the triggered release pipeline moving the trained model to staging.
- Explains the automated process of deploying the model to production after staging testing.
07:59 🚀 *Model Deployment and Testing*
- Verifies the successful deployment of the model in production.
- Illustrates the end-to-end automation, from code change to production deployment.
08:27 🤔 *Reflection and Audience Engagement*
- Reflects on the journey of the MLOps series and encourages audience engagement.
- Invites feedback on content and requests suggestions for future topics.
09:51 🙏 *Gratitude and Closing*
- Expresses gratitude for the audience's time and participation in the series.
- Reiterates the purpose of sharing knowledge and impacting the community positively.

Made with HARPA AI

[pankajverma]

This is great series. Thanks for sharing 👍👍

[jakobp.]

I don't understand how you only have a couple hundred of views on those videos. It can be hard getting into devops especially when you utilize azure ml. Your video is definetly a great starting point. Another video on how to deploy ML Models with devops using docker (without az ml) would be amazing too :)
One thing I realized was, that automatically deploying an AKS will fail on the second try, since it already exists (currently there is no "exist_ok" parameter ), but setting "continue on error" worked fine for me.

Btw, I love your introduction music :D

[dataengineernotes]

WOW, superb, , this is exactly what I was looking for. Thank you.

[prithiveeramalingam]

Thank U so much MG for this awesome playlist.

[ranveersingh-ldeb]

Thank you so much, MG for this awesome playlist. Good Job

[jayantisingh]

If you could now start making series on LLMops that would be very kind of you🙂

[shubhamification]

Hi MG, this series was really helpful. Also wanted to understand how can we add continuous model & data drift monitoring in this pipeline so that retraining can be retriggered automatically.

[macmohan]

Very informative, only concern is how and where do you get those online scripts which you use in the bash task. Good initiative and wonderful.

[sumangupta]

Thank you for the series. It is very good. I want to understand how the pipeline can be triggered based on the model performance like accuracy, f1 score etc. I mean any degradation in this matrix will trigger the model training on new data. How we can ensure that.

[vigneshnagaraj]

Is it possible for you to show the demo of automated ml deployment in azure devops using python instead of Azure CLI?

[anuragkhare]

Really Great series. I would say best reference on youtube for Azure MLOps. Are you planning to create video on continuous learning?

[jatinpawar]

Thank you sir, that's helped me a lot

[ccgce]

I want to know when and how do you increment the compute target aks to aks-02, aks-03 and insurance service also from insurance-service-prod to insurance-service-prod-01, insurance-service-prod-02 and so on...