Azure AI Portfolio
In this blog, I will introduce you to the different types of Artificial Intelligence solutions and services that are available to you in Microsoft Azure. It can be quite overwhelming at first, but the good news is that there is a fitting proposition for each type of data scientist.
I summarized the most common services in one view below to guide you:
💡
This image does not comprise everything, if you want to dive deeper into the topics of large-scale infrastructures with hyper-clusters or hybrid deployments, I advise you to start here: AI at Scale — Microsoft Research. Next to AI on Azure, Microsoft also offers AI capabilities through other media: e.g. Power Platform, and Bing.
To spice it up a bit, let’s take another approach and go through the overview following the structure of how a data scientist could approach a project (ignoring all the steps before the fun part 😉).
After the business understanding and data gathering combined with preparation pipelines, the data scientist’s train of thought could go as follows:
Is there an existing service that can solve my problem?
First, before getting your hands dirty and immediately jumping into coding. It is a good idea to investigate possible services that are built for your issue. These can bring many benefits, such as avoiding the waste of training resources and removing the burden of maintaining the models. With this in mind, have a look at the scenario-specific / applied AI services:
Another option in this stage is to have a look at Azure Cognitive Services if you want to embed cognitive abilities. In other words, if you need a model to speak, see, hear, understand, search or make decisions, then definitely consult these. To get a clear understanding of the broad variety of features behind these services, visit Cognitive Services — APIs for AI Solutions | Microsoft Azure. I pasted some screenshots together for an overview below:
Is there a pretrained model for my problem or should I train one myself?
Of course, there are many situations in which the perfect service will not exist or match your needs. In these situations, it can be tempting to start coding. Pause and think. It cannot be that you are the only one with this problem, can it? Often, this is a valid concern. When you are struggling with these mixed feelings, start looking for the golden mean and reuse what already exists. Many collections of pre-trained state-of-the-art models can be found online and implemented or used as a baseline for further optimizations.
Finally, if there is no prebuilt AI available, you will need to take matters into your own hands.
In this case, it is time to start having a look at the broad ML Platform below (again, there is more on the market than what is shown here). There are possibilities for all skill levels and any kind of machine learning. Furthermore — even though I am not diving into the topic in this blog — the platform allows you to cover the end-to-end machine learning lifecycle. Think of deployment, explainability, MLOps, etc. All while supporting open-source frameworks and languages.
For instance, if you prefer a no-code approach, try out the Automated Machine Learning features in Azure Machine Learning (AML) or Azure Databricks. If low code is more your cup of tea, build your models with the drag-and-drop designer. Do you feel the urge to code? No worries, you could start working in notebooks through, for example, ML in Azure Synapse Analytics Spark pools, AML, Azure Databricks, or Azure Data Science Virtual Machines (DSVM). For more options and details, visit Microsoft machine learning products — Azure Architecture Center | Microsoft Learn.
In this blog, I tried to provide an overview and summarize the most common AI solutions on Microsoft Azure. Needless to say, the data science story does not end here. After building and training models, important follow-up questions could be: How can the solution be explained? How can I deploy this? How can I maintain this solution? And so on.
If you liked this, make sure to visit the official page as well: Azure AI Platform — Artificial Intelligence Service | Microsoft Azure.