Bird’s-Eye AI | Bhaskar Raghunathan at Takvaviya

DISCLAIMER: Below is an AI generated transcript. There could be a few typos but it should be at least 90% accurate. Watch video or listen to the podcast for the full experience!

Bhaskar Raghunathan 0:00
Now that the image data was not greatly structure, it was primarily tons and tons of images. But we wanted to start deciphering information from the images, not just mere photographs and video graphs, can we start looking at more information directly from the images itself?

Alexander Ferguson 0:21
Technologists, entrepreneurs and science fiction writers all seem to agree that for better or worse, drones will be a ubiquitous part of our future. Among the many possible uses drones ability to capture images seems especially potent, but perhaps still yet unrealized. With a background in image processing and data visualization, Bhaskar Raghunathan wondered how a drone imagery could be put to better practical use. And this question led him to found Takvaviya, a company that analyzes image data and translate that information to useful insights. They currently focus on agriculture and solar sectors. But the potential seems vast. In this episode of UpTech Report. Bhaskar tells us the story of his company. And that was technology works. Best and I’m excited to be with you learn more about the journey that you’ve been on. To start us off, can you describe your company? In about five seconds? What would you say?

Bhaskar Raghunathan 1:15
A growing tech company using machine learning,

Alexander Ferguson 1:18
growing tech company using machine learning and the problem that you’ve really honed in on solving what is that

Bhaskar Raghunathan 1:25
we’re looking to solve fundamental asset management problems for large scale industrial applications. We enable energy infrastructure agriculture, we’re looking at how image data can actually enhance field operations. And we are converting a lot of unstructured image data to structured metrics that will actually deliver insights on how your field operations are improving so that your operations could be more efficient than what they are currently now.

Alexander Ferguson 1:55
For you, is this your your first startup running? Yeah,

Bhaskar Raghunathan 1:59
this is my first startup. Yep.

Alexander Ferguson 2:00
Okay. So and it’s been five years, I would

Bhaskar Raghunathan 2:04
say four years we complete for tomorrow.

Alexander Ferguson 2:08
Oh, wow. All right. We’ll have it almost anniversary. Yeah, yeah. of that growth. So this realizing the the problem of large scale assets, we have tons of particularly you said mentioned solar. And what was the other one? Agriculture, agriculture, tons of assets, being able to use data, aerial data, to understand what our problems and being able to fix them? How did how did this really hone in at the beginning and the past four or five years now of what the how you started and realize we got to solve the solution to now where you are today.

Bhaskar Raghunathan 2:43
So the fundamental strategy builds on how we could use large scale image data, the story of the startup, we are passionate engineers, I started it and I had a few interns working with me, we wanted to understand how a quadcopter, a very simple version of a drone could actually help people in solving some fundamental problems. So when we started building a quadcopter, we started looking at, we were able to get a prototype within three to six months of time. And we were able to successfully patent it as well. And once we got that done, we said, Okay, why don’t we put to real time applications. And at that was the time when people were looking at how data can actually drive a lot of decision making. And it was all image data right now. But the image data was not greatly structured. It was primarily tons and tons of images. But we wanted to start desiring information from the images, not just mere photographs and video graphs, can you start looking at more information directly from the images itself?

Alexander Ferguson 3:41
We talked before here about the fact that, you know, you’re not going to focus on hey, we’re on the drone side, right? of flight, you’ll assist with it, it helped me understand and give a use case of, Alright, where do you really heavily play on? And then how do you work with others to even license out your technology? Right,

Bhaskar Raghunathan 4:03
so the core strategy that we have is our image processing and machine learning algorithms. That’s completely AI driven, where that in itself is a plug and play module where we operate with drone operators, we operate with flight planners, we operate with asset owners, where they do the data acquisition themselves, they know how their asset is actually behaving, we bring in the inside part, to the data that they collect. We tell them that your asset is actually doing this well. You need some more things that need to be done. You need to improve process, your your inventory is being controlled this way. And that’s our key focus. Now, there are lots of fly planners, there are lots of pilots who actually collect a lot of data. We provide the information to the data that they collect in a very seamless format, and that’s where the partnerships are happening. So we’re looking at partners who have flight management systems, right? They might have 10s of drones or 20 drones in there. system itself that might be flying all over the globe, or different locations in the US. And they would want they have a platform to monitor all their flights. But do they have a very seamless connected platform inside those flight management systems to tell what is happening to each and every single asset? We’re launching our subscription based model. For a couple of products, this is in the solar PV industry, we will be launching them early May the supposed to launch it sometime back, but the current scenario is we’re looking for the right time, we have a product in place, we want to understand what would be the right time to get into the space because we already offer a lot of services. So tell

Alexander Ferguson 5:42
me more about like, let’s dive in a bit more of why is it different? Like the concept itself is not unique. But tell me you, you talked about really going into the different layers?

Bhaskar Raghunathan 5:52
Right, so So individually, so let me take one particular example, right, I’ll take an example of solar PV, and then probably extrapolate it to what we’re doing in agriculture as well. Some of the key things that we have done over the last few years is to understand where the exact use case of our algorithms is. So when I say use case, for algorithms, this, we have a lot of data that we have collected based on the process that we have already done. And we are running through machine learning algorithms that are actually able to reduce the turnaround time for processing all this information. Right. So the image supply chain of how the data getting collected, is getting transformed into metrics, that’s going to get the addition being done. That image supply chain is what we are disrupting in a way, right? Because you lot of image data gets a lot of time to process. And there are large scale JS industries that are working in this space to crunch all this data into meaningful numbers. Our algorithms are really quick and really good in compressing all this information into meaningful metrics in a very short span of

Alexander Ferguson 6:56
time, would you say? Because you do have your your multiple places around the globe that that you’re able to write serving right now. But I’m curious that that that imagery, does it look different? And is the machine learning algorithm able to understand the differences?

Bhaskar Raghunathan 7:11
Yes, so. So there is definitely a key role in how geography plays in collecting the data and the type of datasets that we have to give you a good picture. We are right now handling some of the assets in Texas, we are also handling some of the assets in India in Rajasthan, right. So in Texas, the place that we are operating, it’s a greenfield on top where a solar PV asset is getting built. Whereas in India, it’s on top of a desert, right. So there itself, you know, you have a huge difference. Right now, our algorithms are not just look at the background on or the terrain on which it’s getting set up, we look at the individual assets that is getting built, meaning the structures that go into solar PV, it’s going to be the same vertical pose, it’s going to be the same set of modules, it’s going to be the same set of model modeling structures,

Alexander Ferguson 8:01
passing faster, fascinating. And this this technology where you really hone in on on that the image processing and machine learning you you are now licensing it to others this this program. So if other we talked about earlier, drone operators, whatever, they could even be using your platform, you’re looking to do

Bhaskar Raghunathan 8:17
that. Exactly, exactly. So some of the key things that we’re looking at partnerships is drone operators who actually have their own drones to fly and collect data can directly dump the data directly to our platform, and use the algorithms and the and the analytics tool for themselves, and then display it out as their tool so that they could white label the stuff and then display it as their product. So the licensing is what we are very focused on. Because we know that there are quite a lot of pilots who have a lot of drones, and who actually want to show meaningful information from the data that they collect, rather than we have photos and videos.

Alexander Ferguson 8:52
What’s the future for your company look like in the near term and long term? So the next year, what are you working on? What are you seeing that you’re heading to? And then the long term or 10 years, where are you headed?

Bhaskar Raghunathan 9:01
Right? Okay, so the this year and the next year, we’re just focusing on expanding our territory, expanding our business in these domains in the US, primarily in the space of renewable energy, infrastructure, and agriculture. That’s been our key focus. In the next five years, our focus is to get into build a much larger tech company that’s completely image data driven. And that’s going to be largely AI driven. When I say AI driven, we have not forayed into spaces of asset management at a much lower level we want to get into industries. We want to get into media and communications, where there are multi various use cases that can be driven based on well annotated image data set. Now our algorithms are completely focused on image data, this slowly delving into the audio space as well. We’re looking at solving some completely different set of problems. We’re looking at mental health how video and image data could solve problems in mental health. Can we decipher a person’s facial expression? When he is actually speaking? Right? Can we tag it to an emoji when he’s actually speaking and and identify what’s going through his brain right now because we deal with raw sets of image and video data. Right now we know what all image and video data can do. So we want to understand build our algorithms on a much higher and a better pedestrian, so that we will be able to solve much more complicated problems in the next three to five years.

Alexander Ferguson 10:34
Be sure to check out part two of my conversation with NASCAR in which he offers some important lessons in developing a product and discovering its real world applications.