'How a Cloud-Agnostic Approach Simplified Complex Data Analytics for Product Management'

Sure. Um, Well, this, this kind of goes back to a slightly related thing, but it's a simpler problem than the entire machine learning pipeline. But I've noticed that um in, in smaller organizations, uh, there's often a need to be able to um quickly like suck in some data, um, and, and then just get great uh descriptive analytics on it. And um one thing that um I sort of was forced to think through in my organization because we, we, uh we're working with different types of cloud systems. So we're working with Amazon, Amazon Web Services, with, with Google Cloud platform, with Digital Ocean, with, with other things. Um, Uh, what, what is a, a very simple way that a person could get stood up with all these, these things, like how can I pull data, um, and then essentially just look at it very quickly, and that's cloud agnostic. And so, um, the simple solution was to identify, um, You know, the, uh, Uh, You, you know, The, the reference spec basically for doing this, uh, in a cloud agnostic way, and then make it so that um someone could uh deploy that anywhere to any cloud, you know. So, uh, a simple, a simple database that feeds into um a, a very user friendly, um, Uh, kind of dashboarding solution that's open source. So pretty much all the pieces are, are free and open source so it costs you nothing, uh, uh, other than like the, the cost of the, the CPU, um, which is, which is very lightweight if you're trying to do something quick. Um, and I, I share that with my team. And at first, you know, I, I, I worried that it was too simple, you know, because often, you know, people want to do like the most complex thing and do that, but uh, I, we ended up using it quite frequently because it was so simple, you know, people would have a need, like internally, they would have a need to, you know, spin up, uh, you know, some, some graphics that we're pulling from the data. So people use it all the time, uh, just because it was, it was so simple and it was free, and it was like just, just there for you to use. So is that reference spec uh for, for doing that and then kind of automated it with, you know, some of that terraform stuff that we're talking about earlier.

Uh, well, I had been evaluating the, uh, the offerings from different, um, Different cloud providers. And we were having a bunch of back and forth debates about lock-in and being cloud agnostic and, you know, uh, there's sort of an ease of, of entry of, of just picking one cloud provider and just rolling with all their stuff. But then, um, Like I said, we, we wanted something that could be applied anywhere. And um I think that, that's what led to it is just looking at what the different offerings were, doing my own kind of research about the different um uh dashboarding tools, and then focusing on um ease of use and ease of like startup. Like what is, what are, what are the easiest set of things to set up from the beginning? Um, cause there, there's like a ton of different, uh, dashboarding tool, you know, re dash, and I mean, there's, there's like so many different ones. And it's, I, I just looked at across all and looked at all these reviews, we tried some out. And then, you know, we just said, what is the simplest possible thing that we could carry to any cloud-based environment? Um, And, and then we just specked it out and then, uh, and yeah, we, we, we use it for, for many different things, um, uh, internally and at uh portfolio companies. Uh, and, and the other kind of, uh, constraint that I, I, I, I try to work within was, even though it was simple, uh, all the pieces needed to be scalable, so that you could, it could, if, if you wanted to scale it up to something that was larger, you could. Um, and, uh, I think that is also what made it attractive.

Um, what wasn't a product as much as a sort of a tool, I mean. A specification for a set of tools and how they would interact with each other. Uh, you wanna see like a toolkit with that show, where, where I was kind of saying, you know, if, if you, if you need to kind of quickly spin up something, then use these tools in this way connected like this. And um, what, like I said, we, we ultimately ended up using it. For, for me, for that experiment framework that I was telling you about, we use it for that. Uh, we used it, um, for one of our portfolio companies, uh, for more complicated, um, you know, uh, series of, of, of data ingestion into a uh like a, a, a data warehouse pipeline that would Resulted in kind of uh KPIs being presented through that same dashboarding system. So it could scale up to larger problems, but at its core, it was very, very, you know, simple uh set of tools that, that you can start from.

Don't focus on the fanciest thing. That you could build. Um, because when you are in an experimental mindset, and you want to be able to move really fast, the simplest, quickest, easiest thing is going to allow you to get your experiments started much faster. And, um, and then, you know, at some later point, you could decide, you know, you want to To replace pieces with more fancier things. Can I