Relationships in Pharma Manufacturing Data

Welcome to this video on relationships in manufacturing. If you do not understand why relationships are important in manufacturing, or you have never used it, then this would be a great video for you to truly understand what is it that we mean by when we talk about relationships and manufacturing. From when it comes to manufacturing, we are primarily talking about pharmaceutical manufacturing, but let's talk a little bit more about in terms of what we mean by relationships.

We are talking about relationships in data. While people have a feeling that they have solved a data problem or they have a path to solving, let's say, data problems so that they can truly get the insights from manufacturing data, it has been very difficult and most people have not been able to solve that problem to their satisfaction. This particular video will help you in understanding the basic concepts around relationships, knowledge, knowledge graphs, so that you can understand a whole different way of solving these problems so that you can truly harness the insights that comes with any data, and in our case, it's manufacturing data.

In order to understand relationships a little better, let me take you back a little bit to the 1960s to the Apollo program and the Saturn rocket. Very complex system, lots of different parts. All of these parts are, not all, but a lot of these parts are manufactured through multiple different vendors to a particular specification.

They come together, they fit together, and then the rocket works as it's supposed to work, or it does not. Now, NASA really wanted to understand how to properly understand the connection or the relationship between all of these components and subcomponents, because let's say if a particular motor is not working well, then you need to understand, okay, what does that mean for that particular subsystem, and when that subsystem does not work well, what does it really mean for the overall rocket's performance, right? And you need to understand this so that you can take corrective actions. Now, in order to solve this problem, they were truly looking for a way where it's easy to understand the relationships and to save the relationships in a way that you can truly analyze it, right? And they partnered with IBM to create something called an IMS database.

It's a hierarchical database. In this hierarchical database, it not only takes care of the data, but it also accounts for the relationships. And in my mind, I think this is one of the first few knowledge graphs that ever were created, and it was created pretty successfully because that became the way of managing these complex relationships in a really complex system.

Fast forward a little bit to the time of, let's say, you know, search on the internet, and AltaVista and Yahoo, these were the leading search engines of their time. And the way they handled search was essentially understanding the websites and the webpages that were out there, understanding the content, and then trying to relate that with the search that is coming in on which ones are more important and which ones are less important. Comes in Google.

Google comes in right after that. And once they started giving in search results, people never went back to AltaVista and Yahoo because the Google search results were so much better. Now, what was different in the Google search results is while they also took the data that is sitting inside the webpages and the websites, they also focused on the relationships between these websites.

So, if there are links between one site to another, and then from there to someone else, those links, those relationships, they use that in their page rank algorithm to try and marry which sites and webpages would be more relevant to the information that people are looking for. So, here again, relationship became the way of truly solving a particular problem where people wanted to build out, let's say, insights. Now, coming to the pharmaceutical manufacturing, there again, if you think about it, the pharmaceutical manufacturing in some cases takes over a very long period of time.

And it happens in multiple different, let's say, steps or stages. Sometimes from, let's say, all the way from the raw materials to actual drug going into a pill or into an injectable, it might be in a lapse time of seven to nine months, maybe longer. And there are multiple steps that have happened in between to create, let's say, parts of the, let's say, semi-finished raw materials that go into another process where it gets worked on and then something else comes out of it till your finally drug product comes out.

And there are multiple different things that are happening along the way. And let's say it's in this particular world when you are, let's say, two or three steps removed from something that has happened in the past, it becomes extremely difficult to know why things are happening in a suboptimal fashion right now. And if you really think about this, if you truly created the relationship between all the activities and the steps that have happened right from the inception to basically the final drug product going out, then it becomes an easy, easy problem to solve or easier problem to solve in terms of using the power of relationships to truly understand exactly what the dependencies are across all the different components and subcomponents of this very complex manufacturing process.

This is the knowledge graph in a manufacturing, in a pharma manufacturing, we call it a genealogy, a bad genealogy, which truly relates to how all the things are coming together in order to create the final finished product. Now, another important thing in order to basically understand this particular point in greater detail is the true value of your data is essentially more in the relationship space than in the data itself. Now, if you look at natural biology, the human brain has about 86 billion neurons, but it has more than 100 trillion synapses, right? So you can think of neurons as those elements that basically store the data, and the synapses are the relationships between the data.

So even biology knows that it's the relationship between the data points that basically are more important when it comes to deriving insights or building insights from raw data. And this thing is also showing up in everything that we are seeing in the innovation in Gen-AI.

The relationship between all the different pieces of data, when stored appropriately, then the Gen-AI can actually take out insights, build insights from it in a really, really smart way, where a lot of it can be really useful.

We also know that some of it can be wrong, because it can, it has a tendency to hallucinate. Now, this is where the knowledge graph that we built becomes extremely important, because here we know or we have defined a way of building out your manufacturing data knowledge graph in a very safe GXP-relevant way, which can then be used to truly build out all the different insights that you have always wanted to build out. And the traditional ways of solving this, of, let's say, bringing our data from multiple disparate systems into a data lake, it does not necessarily get you there, because all you have done is you have taken disparate pieces of data from multiple sources, and you have basically arranged it in one place, where you still have disparate sources of data just organized in one particular place.

It does help you a little bit, now that you have one place to go and get your data, but it doesn't solve your problem of being able to truly derive insights from it, right? Hope this was useful. Hope this gives you an idea of what we are talking about when we talk about relationships in manufacturing intelligence. In the next video that we are going to publish, we'll go a little bit deeper in how we build these knowledge graphs from a manufacturing data that can then basically become the core data asset on which you can build out multiple different kinds of insights.

So, looking forward to seeing you again in that video. Thank you so much.