Business Intelligence and SCM
This web page is dedicated to BI and AI programming. A lot of topics are taken from my domain supply chain management, but there might be topics from other areas as well. The choice of programming languages and frame works is free (Power BI/DAX, Python, PyTorch, C#, Azure, …). The aspired solution dictates framework and programming. If interesting, the solutions are embedded in a framework. Going away from the 101 standard script-like solutions, in the real world the requirements are different:
| Programming is necessary: |
| Why programming at all? Because the more possibilities, the more ideas, the more business requirements. The capabilities of BI/AI systems will shift upwards as well. However, the development of configurable of-the-shelf products will never keep pace with emerging ideas that will need creative programming.. But it does not stop there: with the advent of AI solutions that work reliably to provide decision support, BI and AI solutions will be blended. You will have to program in your BI and AI frameworks. |
| Programming is programming in a framework: |
| A common narrative is the 101-approach that lets you successfully program standalone script-like solution for a specific problem. In the real world this solution is embedded in a universe of infrastructure that guarantees functionality, integrity and correctness of data and results. It is practically impossible to build up this infrastructure from scratch, nor it is in any way desirable. |
Long story short: solutions are always embedded in a process. All examples and solutions presented here are meant to support this idea. As soon as you leverage existing features in your infrastructure everything is fun, because it works in the real world.
The other development is, that AI will take over the world. Everything what is conventional wisdom and is not innovative will be replaced with time. This is the same situation as it was in factories, where counting screws is now long obsolete. Some blogs will tackle the problem, if AI can replace human decsion making to some extent. Ths already done in planning software and the like. The focus here are the inner workings of those models and how to program them.
If this is true, then AI relevant techniques for big data are applicable: vectorization and parallel computing. Big data requires efficient computing. We will come across setups on the GPU quite naturally. Again, if you go to an organizational level, distributed computing will be sometimes necessary.
As mentioned, there will be a lot of examples from supply chain management, as I work in this field for a long time. It is a good field for interesting problems: a supply chain has a lot of actors and their interaction is, gently expressed, noisy. You can challenge yourself, from an isolated classfication task in a warehouse (e.g. product recognition in a warehouse), up to a digital network twin for an entire supply chain.
These will be a loose series of blogs as time available permits. Often, there might be a series of blogs, where I start easy and hope to develop the subject. If you are interested, please leave a comment to discuss, critizise or praise the contents!