21 minutes full of #googlecloud Vertex AI Pipelines. Basics, best practices, code, and everything needed when putting #machinelearning pipelines into production. https://t.co/wo7aVbPzUI If you enjoyed this video, please subscribe to the channel ❤️.

Coding GenAI Applications

Testing upcoming livestream

Google Vertex AI model versioning support is now part of the SDK (3 hours ago 1.15.0 was released)

(I can confidently say I highly underestimated the effort needed to create YouTube videos 😅. Setting up proper audio is a challenge itself, cutting, processing 4K videos, not talking about the efforts needed for the recording)

Back in the days, at a time when #MachineLearning was not all over the place. A large number of Cloud Composer workflows got created. This article is for you if your previously only data-related workflows are migrating more into an ML workflow. https://t.co/7bbcE805LF

Curious about innovating faster using ML on #AWS? Join our event at the AI Campus in Berlin on May 10th at 5:30pm and get a deep dive into #ML services and hear from @creatext_ai how they are using @awscloud to better serve their customers. Register: https://t.co/xIqgaaaa6F

13.500.000.000 Euros the EU has paid for Russian coal, oil, and gas since the Kremlin’s war of aggression on Ukraine began on February 24th! Counter: https://t.co/VTDVvyKZ3d

Great to see so many people joined in person and remote the @GDGBerlin With speaker from @doitint @24metrics and @Merantix

We are creating an IT army. We need digital talents. All operational tasks will be given here: https://t.co/Ie4ESfxoSn. There will be tasks for everyone. We continue to fight on the cyber front. The first task is on the channel for cyber specialists.

#SWIFT Shame on you @OlafScholz #olafshameyourself

The last step needed is our compiler. The compiler takes our pipeline and creates a pipeline specification as JSON. That’s all that we need to run the pipeline. We can now use Google Vertex AI Pipeline and run the pipeline using the API or UI. See you tomorrow with more...

The pipeline itself contains the components we created. You can see the output from the first component is the input for the second component.

A Vertex AI Pipeline consists of multiple steps, where each step is defined by one component. The component contains the code that this pipeline step should perform. What we see here is called a function-based component. It’s the simplest one, we just write a python function.

Because we are implementing a Kubeflow Pipeline we need to import the required Kubeflow Pipeline Modules. We go over each of those Modules

Things are simple until you make them complicated. Exactly because of that, we start with the most simple pipeline possible. Don’t worry we get to the best practices and deep dive topics very soon.

ML Pipelines are there to connect the various steps of your ML solution. Kubeflow is built to run on top of Kubernetes. And Running Kubernetes Cluster can be challenging and time-intensive. That’s why Google introduced Vertex AI Pipeline a product to run pipelines serverless.

Let us start with a quick introduction to ML pipelines and how Kubeflow and TensorFlow Extended are related to Vertex AI Pipeline If you’re not familiar with Kubeflow or TFX don’t worry too much. Everything you need to know we cover during this thread.

Make sure to comment with your questions. See you tomorrow with additional content.

I am going to start this thread with a bold statement. Machine Learning teams don’t need Kubernetes. Thanks to @googlecloud our daily work with machine learning pipelines got way easier. Follow the thread if you want to know how that works.