Preparing Data for AI Apps – Techstrong AI Podcast EP24

Hello and welcome to the Techstrong AI Podcast. I'm Amanda Ani, and I'm so excited to be here today with Krishna Ian, and she is the co-founder of comprise. How are you doing today?

Doing great, Amanda. So can you share a little bit about the company? What does comprise do?

What services are offered? Sure. Um, comprise is, uh, a company that we created, uh, a few years back, um, to address the problems with unstructured data.

Uh, so, um, you know, this is our third company together for me and my CO two co-founders. Um, and a lot of our customers from our last two companies came and told us a few years ago that a big problem they're seeing is that they're drowning in unstructured data. Things like files, videos, images, you know, genomics documents, you know, all those kinds of things suddenly is becoming larger and larger.

And today 90% of all the data is unstructured, but most of the systems, most of the storage and the data management systems were really designed for databases for structured data. So how do you handle all this unstructured data in the cloud, you know, in your data centers, how do you know what you have? How do you ma move it to the right place at the right time?

And how do you index all of it? Those are the problems we solve at comprise. So it sounds then like you are the person to speak with about today's topic, which is preparing data for use in AI apps and tools.

So when it comes to this subject, what is step one for companies during this process? Well, you know, I, the, the first thing I think to, uh, realize is that, um, AI, you know, is, uh, really good, uh, in looking at large swats of data and, you know, gleaning something from it. And the better, more data you train it on, the better it's going to be at recognizing patterns.

So AI and data go hand in hand. So AI is better when you give it better data. Uh, so that's why preparing data for AI is very important.

Um, so when an organization is thinking about preparing data for ai, uh, you know, they have to look at their data first. And, uh, there are different techniques for the structured data and for the unstructured data because structured data probably is already in a database or it's in several spreadsheets or parquet files. So it has some structure, you know, it's either already has a data piece or it has some tables that you can then index into a catalog.

Unstructured data on the other hand, has no such structure. So the first thing you have to do for unstructured data is find some way to create an index of the unstructured data. So, uh, regardless of what data you have, building an index of it or a catalog of it would be the first step.

Okay. So let's talk a little bit more about unstructured data, that that is a, a big issue you mentioned. And what are some of the roadblocks that you see business leaders coming up against when it comes to that?

Well, you know, unstructured data, Amanda, like, I don't know. Um, you know, if you take a lot of photos on your phone, uh, for example, uh, you know, uh, most of us take a lot of pictures, a lot of videos, but if I asked you how many pictures are on your phone or how many pictures did you look at in the last month, you probably wouldn't know, and you probably only looked at a couple of the pictures that you took, but if you're like me, you probably didn't delete anything. Right?

Uh, and, and that's what most organizations are like as well. You know, they have a lot of data that a lot of users, um, accumulated, a lot of applications generated. The data is sitting in many, many places.

Nobody really knows how much data they actually have. Uh, and so the biggest challenge when it comes to data is first getting analytics on the data. You know, getting to understand how much data you have, how fast is it growing, what kind of data is it, you know, who is actually using it, what is valuable, what is obsolete?

Getting all that information, uh, is the first step to even starting any process. And what is an issue that comes up during this process as far as the staff communication? Because I know one issue I hear a lot is communication between departments and making sure everybody's on board, understanding the process.

So what are some tips there? What are some things that you've heard? Yeah, you know, um, uh, especially when you're talking about, um, AI and data, uh, you know, one of the big concerns is around, you know, data security and data privacy and data governance because, uh, you know, a lot of, especially if you're using an a third party AI solution, especially generative ai, like you're using a large learning model, you have to be really careful when you give it corporate data because it's getting trained on your data and you have to make sure that, you know, there's the right legal agreement.

So, you know, your data is not inadvertent getting shared in the public domain. You also have to make sure, you know, things like, you know, PII or sensitive information is not there when it's where it's not supposed to be. So when it, when it comes to employees and employees use of ai, I think having really good governance principles and guidelines and some way to automate the auditing of, you know, what data gets used for AI is going to become extremely important.

Uh, because we still don't even know all the security ramifications and legal ramifications, uh, of the use of AI yet. Uh, so, so organizations have to be extremely diligent upfront in putting the right, um, sort of, uh, policies in place and automation to track those policies. This technology is advancing so quickly.

What are some use cases or some examples that you can share with this technology implementation? Well, you know, you could think about it. Uh, a lot of times I think that, uh, an easy way maybe to think about ai, uh, and maybe our children actually already know this.

'cause you know, if you have a a, a teenager, they're probably already using, uh, some ai. They're, when they're reading long passages or when they're trying to understand something or when they're doing research. But take of AI as it could be your, your personal assistant, it could be your assistant looking through tunnels and tons of data and gleaning some insight for you.

So if you are a hospital and you have to look at lots of images to find all the ones that have a particular mutation because you wanna do further research on it, uh, and you wanna identify like where that is, instead of having people go and do that work, which is very tedious and time consuming, you could train AI to do that work for you. Uh, and so I think the simplest form of AI will, will start with automating a lot of manual tasks, uh, which requires some, uh, intelligence to recognize things so it's not just truly manual. There's some intelligence involved, but that intelligence can be learned.

Um, so that's a good example. I think. So image processing, you know, uh, video search, audio search, um, you know, chat bots, um, um, you know, basic, um, data processing.

I think these are all robotics. These are many areas where we will start seeing applications of AI very quickly. Uh, Yes, I've heard a lot about the AI and robotics and many interesting use cases there.

Yeah, absolutely. Because, you know, you think about, uh, you know, robotics, uh, you know, you can automate when we, when we think robots, we think a humanoid robot that walks like a person and does all uh, everything like a person, but there's a lot of robots that don't need to go that far. I mean, they could just be part of an assembly, uh, line and they could be automating processes there, or they could be doing basic cleaning or they could be going to dangerous places where humans can't go.

You know, you think about space, you know, exploring space and you know, we call survive and space and the radiation of space, but maybe that's a good place for robots to go and prepare a settlement that we could then go and live it at. So those are the kinds of amazing things that are now possible. Yes, I could see it being used for a lot of rescue mis missions where other people are putting their lives on the line to try to rescue someone, but they instead could use some sort of robot AI technology.

Yeah, exactly. Yeah. So if there was one key takeaway that you could leave our audience with today, what would that be?

I would say that, you know, you have to start somewhere. I mean, this technology's advancing rapidly and sometimes it can feel overwhelming, um, because you know, every day something new is coming up. But the simplest way is to just start by, you know, getting some analytics on your data now start with a simple use case and just try it out.

Uh, and then you get a much better feel for what is possible. Alright, Well thank you for coming in on our show and sharing your insights with us today. Thank you so much Mano.

And thanks to our audience for staying tuned. Have a wonderful week.