Azure OpenAI is at the moment on a per usecase submission basis for making use of it. You can apply for a spot through an application form if you like to use the services. I have played a bit with the services but no significant knowledge about the service specifics yet.

@Dave Ebbelaar do you use the clients Azure tenant and make requests on the per usecase scenario or do you have a partner deal with Microsoft to circumvent this procedure?

Normally in a corporate environment you would be cooperating with a data engineer or a data engineering team which task is to maintain and provide data to the data analist and data scientist. This data would ideally reside within a Data lake/Data Lakehouse architecture. These kind of architectures enable your point like versioning, history, access management etc. My key point in autodidact learning is exactly about not knowing these best practices because the focus of learning is mostly on the data science/analist aspects. I'm not saying that this is wrong but I think dat the awareness of the whole data engineering trade should at least be in the back of the mind of the data scientist so that he is at least aware of the best practices and methodologies used for data storage and governance. This way the cooperation with the data engineers should be better. For example: In a past project I was working for a large telecom company which asked me to optimize some queries. One query in particular for calculating all phone costs on a event level per client ran for more than 4 weeks per run (It was crunching Petabytes of data) . Because of this length the billing process was not able to be on time. The query I was looking at was built by a data analyst without knowledge of the best practices for querying such huge amounts of data. After I finished optimizing the query statement the query ran within 24 hours per query run. This example was very obvious because of the huge amount of data processed. But most of the time it's not so obvious that bad query choices have a big impact. A frequent solution for fixing those smaller issues is to upscale the processing power of the system instead of fixing the root cause of the problem.

Hello, I'm Michael from the Netherlands and I am a Business Intelligence Consultant working for 13 years in the business. I'm used to working with data on a corporate level with datawarehouses or data lakehouses as a foundation for reporting and dashboards. I want to extend my knowledge towards machine learning and artificial intelligence.