In episode 03 of the Let’s Connect! Podcast, Muddu Sudhakar of Aisera joins us to talk about Unsupervised AI and how to leverage algorithmic intelligence to create IoT decision trees.

Muddu Sudhakar is a successful Entrepreneur, Executive and Investor. He is the CEO and Investor of AiSERA the industry’s first proactive, personalized, and predictive AI Service Management (AISM) solution that is purpose-built to automate tasks and actions for IT, HR, Facilities, and Customer Service. Muddu has deep Product, technology and GTM experience and knowledge on enterprise markets such as Cloud, SaaS, AI/Machine learning, IoT, Cyber Security, Big Data, Storage and chip/Semiconductors. Muddu has strong operating experience with startups as CEOs (Caspida, Cetas, Kazeon, Sanera, Rio Design) and in public companies as SVP & GM role at likes of ServiceNow, Splunk, VMware, EMC. He is widely published in industry journals and conference proceedings and has more than 40 patents. 

Interested in connecting with Muddu? Reach out to him on Linkedin or Twitter!

Aisera offers the world’s first AI-driven service experience solution that automates operations and support for IT, HR, Sales and Customer Service, making businesses and customers successful by offering consumer-like self-service resolutions to users. Aisera fast tracks the digital transformation journey with user and service behavioral intelligence that drives end-to-end automation of tasks, actions and business processes with unsupervised AI. Aisera is a top-tier, VC-funded startup headquartered in Palo Alto, Calif., and a strategic partner with AWS, Microsoft Azure, Google Cloud, ServiceNow, Atlassian, Zendesk, Datadog, Automation Anywhere, UiPath and Salesforce. Follow Aisera on Twitter!

Key Questions and Topics from this Episode:

0:00 Show Introduction 

1:05 Muddu Sudhakar Introduction

2:45 What is Artificial Intelligence (AI)?

6:30 Decision Tree Automation in Industrial IoT (IIoT)

8:30 What is the Real Impact of AI Algorithmic Science?

12:40 How does AI Learn?

14:15 Intelligence at the Edge of the Network

16:05 The Birth of AIoT

19:00 How Should Companies Get Started in AIoT?


- [Narrator] This is the IoT For All media network.

- [Ken] Hello, friends in IoT. Welcome to Let's Connect, the newest podcast and the IoT For All media network. I am Ken Briodagh, Editorial Director for IoT For All and your host. If you enjoy this episode, please remember to like subscribe, rate, review and comment on all your favorite podcasts and platforms, And to keep up with all the IoT insights you need, visit Before we get into our episode.

- [Ken] The IoT market will surpass one trillion dollars in the next few years. Is your business ready to capitalize on this new and growing trend? Use leverage is powerful IoT solutions development platform to efficiently create turnkey IoT products that you can white label and resell under your own brand. Help your customers increase operational efficiency, improve customer experience or even unlock new revenue streams with IoT. To learn more, go to, that's Now let's connect.

- [Ken] My guest today is Muddu Surakarta CEO of Aisera, and we're gonna talk today about AI and sort of how it works in the real world, what it can actually do and how you should be using it. Muddu, thank you so much for being my guest, welcome to the show.

- [Muddu] Ken, thanks for having me.

- [Ken] Thank you.

- [Muddu] and I hope you are safe and your family safe, and your listeners are continuing to stay safe through this difficult COVID time.

- [Ken] Yeah, it's been a heck of a year, hasn't it? I hope the same for you and yours, and of course I hope all you folks out there listening have been weathering the COVID times safely and well. Muddu in case folks aren't familiar with you or with Aisera and how you guys work within IoT. Can you give us a little bit of your background and where home base is for you?

- [Muddu] Yeah, so I'm the co-founder, CEO of Aisera, and Aisera is based in Palo Alto, I live in Cupertino and Aisera as a company, which is now more than three years old, it's venture backed, which funded by a venture firms in Silicon Valley, like Norwest ventures Through ventures, Menlo ventures, et cetera. And I think in the, in general in the broader AI we play in the broader AI space. The name Aisera is mainly designed around how we automate customer service, service management. I can remember when we started it, there was no COVID. My goal is can AI automate manual deputy to common trust across the spectrum? And obviously IoT starts being up in new challenges, right with 5g. And the 5g is itself 5g cloud and AI is going to hopefully make AI IoT to the next generation.

- [Ken] I think that that is a great place to start because there's a lot of sort of future thinking around AI. And I don't think it's all super connected to what AI is today. I was talking, I was interviewing a couple of years ago Jim Goodnight of SAS. And I asked him during that interview, you know, what is AI? What is, what is he think of as AI? And he said, Oh, AI doesn't exist is not the way people think about it. You know AI is just the way it really exists is data analysis sophisticated algorithmic data analysis. But all it is is looking at the things we know in the way that a computer can and a human isn't super good at. Do you agree with that sort of assessment and and how do you think about what AI really is?

- [Muddu] Look, I don't have any religious thing. I said, I'll tell you my practical view. Look, AI is for sure. It's, AI is not any magic it's math at the end of the day it's about data, math and algorithms, machine learning in AI has evolved to a point where it can learn few things. There are areas that will continue to evolve. Areas where its gotten better is as we have seen it in the areas of image processing, it became really real, right? We have seen AI can do quite a bit of recognizing images. People use it for video. It's very real in the in ad technology you've seen the danger of what Facebook did how using AI, right?

- [Ken] Yeah

- [Muddu] So alternating people's behavior in the new sector. So Facebook, Google and Amazon they're exploding AI for ad technology. I think a lot of AI machine learning happening to services which people talk less and we see this flash trading, right? It's called the Fin Tech space. This whole algorithmic trading is all nothing but AI, right? So algorithms are machines are trading with each other. Sometimes our wealth is impacted by that, right? Our 401 key accounts and our savings accounts and a lot of mutual funds. I think some areas like that have been around already for more than 500 years. And people are already seeing the good and bad of that risk going on. The next level where I see the AI is going to start happening in more realities in sales, marketing, customer service. Right? I think customer service is where I started seeing more of this in 2016 and 17. And that was of service now, right? What started happening is particularly call centers, right? Highly repeatable tasks, right before you wanted to talk to you and being now you, your father, your mother your high school teacher at home you have time to talk to somebody. So if there's a way you can interact with the system and get the job done, right. Highly reputed, because let's say you want to set up your appointments or you want to have somebody to respond to certain things. Somebody has a question about, how to import cash? How do I join? Can support customers? What are Malaysian occupants? They should be able to chat to a system

- [Ken] Yeah.

- [Muddu] But we live agent like human being.

- [Ken] Yeah, yeah, for sure. And, and I think that that, that sort of execution of AI is sort of a, a non endemic one. It's not one that people think of first but I think it's one of the best uses for sort of the the type of AI that we have now where it sort of runs a choice tree kind of thing. And I think it's got a lot of applications, as you said not just in customer service but also for retail and, and, and online shopping. And it's of course gets used there a bunch but also sort of on the industrial side I think that there's a lot to be said for sort of choice tree AI in manufacturing. For instance, if you're, if you're trying to optimize a manual manufacturing line or a a warehousing operation or supply chain in general if you can alter the the sort of decision tree according to to changing, you know, situations, it seems like that's a natural case for AI, right?

- [Muddu] It is. So I think you use a lot of good words like patient trees and make conditions. So I think that, again, the schools of AI that evolved in the last few years, the early days people created these decision trees and some people call them guided flows. Like you saw Google acquired a company called IBM doing IBM, Watson, Salesforce, Einstein, all these are I actually call them some people call them chatbots. Some people call them actually supervise guided flows audition trees. That is, if somebody asks this question what on this decision tree part, right? It is your troubleshooting. Is your, is it an hardware in the device at the edge, is a device not responding? So you could create some kind of addition tree and give that to a system and say to go to that. But what happens with that approach is that works for cases where you already thought through the decision tree, if you've already thought through it, you're almost acting like a God, right?

- [Ken] Yeah.

- [Muddu] You are exactly what your users want to ask and go through it. There's a reason they're there. Sometimes the people going to ask you is not something you can think through. And that's like, think of like how Google created the search for the long exponential tail of what users may ask. You may not know. So same thing, if you want to do that for, for AI I call it unsupervised AI, which is you can't supervise and think through all the decision trees.

- [Ken] Yeah. And that, that's something that I'm fascinated by. And I know I promised we wouldn't get too much into future thinking, but the the real impact for sort of all of this data collection that we do in IoT and all of the cloud storage and, and and also AI algorithmic science that's happening is to me looking for the sort of unknown unknowns, the, the implications of statistical analysis that we aren't thinking of, and that we don't know of because they are perhaps minuit in you know, or they're correlative, not causative. And so we're not seeing how powerful they could be that kind of stuff. I think that there's a lot of advantage to be found in that kind of, sort of unsupervised analysis because I don't know, maybe I'm not arrogant enough but I don't think it's possible for humans to think of all the possibilities.

- [Muddu] Yeah, absolutely. You kind of describe perfectly, right? See if you think of every possibility then you really like acting like a God. So, but if you let the system think through those examples then system also will make a mistake. AI is not perfect to and that's where you use the word data. I think able to use the data to create these decision trees dynamically in, in its brain. So that's where the neural networks comes in. That's where the people call it deep learning part of the AI so if I can use your, let's say if I can take all your podcasts Ken and cannot do text transcript, and actually have AI, learn. I can do unsupervised AI on it. So enable to create edition tree and its own call knowledge graphs. Right?

- [Ken] Yeah.

- [Muddu] So that's where the new unsupervised AI is coming as this idea of knowledge graph, dynamic decision trees in. So you're really not creating addition tree, but you're creating the chain of thought in the algorithmic brain.

- [Ken] So what I'm hearing is folks out there listening this is a big announcement for us. The next show to come up on the IoT For All Media Network will be entirely hosted by AI. We're just going to go through and Muddu going to write it for us. And we're just going to have the AI podcast and our guests are going to come on and talk to our, our algorithmic host.

- [Muddu] That would be nice. But can I think of Ken this whole idea of what we are doing at Aisera today, and when we started Aviznar I was not aware of how much we could accomplish We did. So we actually crawled any content like content could be unstructured content from your web pages. It could be the live agent called Nodes. It could be a PDF document right?

- [Ken] Yeah.

- [Muddu] Any documents and out of that if I can derive what I call the meaning intense because language has something called an intent because in English language you may ask the question in 10 different ways. What is the intent and how many different the intent can be applied. It could be a synonym. It could be, you might use a different phrase. It's called a in Aton sense in AI. So all these things are really important because whenever you bringing back to your IoT, the best example of IoT and industry are two of them. One of them in automotive industry, Tesla. Tesla comes in IoT devices on our freeways,

- [Ken] Yeah.

- [Muddu] Right? on our road. And at home you have a device called Alexa from Amazon.

- [Ken] Yeah.

- [Muddu] These devices are now, if you can make every IoT device has a standard stack and operating system and it has a natural language interface, right? You can actually talk to your Alexa. You can talk to your Tesla. Tesla will tell you when to service your car. So the uptake in the AI to a point they monitor it, they collect the data. They're able to do predictive maintenance in case of Tesla, Alexa, you can talk to it you can interact with it. You can do things. It can only limited things. It can play the music. You can just start my podcast.

- [Ken] Yeah.

- [Muddu] All right?

- [Ken] But, but it's also getting better at, at being predictive. I had not on the Alexa platform in my house. I'm on the competitor, the Google home. But, but it's it very similar operative system and it's getting really good. Like it will send me alerts and say, hey, your furnace seems to be turning on and off pretty frequently. Make sure you check that there isn't a problem. And that's a predictive maintenance thing. It's not sort of determinative. It's not saying here's the problem. It's just saying it's still in the early phases. But it's interesting how it's learning, that kind of pattern.

- [Muddu] Absolutely, I actually, I mean, you nailed it. So it's communicating with you. So you're comfortable talking to it and it is interacting with all your devices, right? So the devices, so that's also missing in the IoT use our data is IOt device is not chatting to one problem with, if you look at why the IT industry did well, the application SAS cloud is if you look at, from Microsoft, they generate a lot of logs. You probably heard the word logs and events and telemedicine.

- [Ken] Of course, Yeah.

- [Muddu] I think this is well-known for generating logs companies like Splunk will entire the company on logs and events, et cetera. Whereas if you look at the IoT devices they tend to be less chatty. They send the gender less data. So even though they're not they're opposite of the big data problem, because it's like, you are a human. If you are a device, Ken I'm just going to use it as if you're an IoT device. And if you don't talk much, I don't know who you are. I can't predict who you are, your behavior.

- [Ken] Right.

- [Muddu] But if you're thought then I can start getting a sense as to who you are and et cetera. Right? So same thing with IoT devices they need to start generating a good enough telemetry data of various points. Because if you just give me one data point, I can't predict about it.

- [Ken] Right. And, and the, the limitation on the IoT side of course is bandwidth is expensive, battery life, you know all these other things. And I think that that's sort of where the the edge of the network is becoming more and more important is being able to to store data and transmit it in the most efficient possible ways or to transmit only the relevant data. That kind of stuff is allowing us to build some really interesting sort of digital twin style models that then we can run the algorithmic AI against and try to create some of these decision tree things. Are you seeing that sort of thing happening in sort of in your work on the, maybe only on the industrial side where they're using the edge data in combination with sort of the big data in the cloud to sort of build these models?

- [Muddu] We do. I think that's an area for more work to happen. I am seeing some but not much. I think on the data, I'll tell you the challenges are as you just rightfully say see IoT devices don't have enough battery capacities, call it storage. So they need to come with an intelligent mechanism to generate enough delimiter to the cloud so that, you know edge is where the actual action is happening. Processing analysis could happen in the cloud like how Google home does it, or Alexa does it right. But able to collect enough data to send it out to some place and creating an IoT device with the right stack to right collect. There's still work to be done. But once the data comes there and as is happening in the cloud, the decision making can happen at the edge. So you push out your algorithm. So, so the edge will decide. So you want to de-centralize addition, meaning it's not you don't need to always talk to your mothership to decide.

- [Ken] Right, right.

- [Muddu] So the algorithm should be compact enough to push the executable into the edge.

- [Ken] And that's, that I think, I don't think we're there yet but I think that that's when we're going to start seeing like real functional AI it's going to stop being sort of a, an experimental nice to have for the enterprise and more of a must have sort of tool rather than a, a toy.

- [Muddu] Awesome. Yeah, I know, it is, it is. I think when that day happens like that will be a big revolution for us right? So I think I'm look, things don't happen always linearly it happen exponentially but industry is definitely going the right direction. The promise of 5g telcos, can do a lot more.

- [Ken] Yeah.

- [Muddu] Don't to pick one area. I think telcos have the network by different telcos, most slow. I think if I was one thing the telco companies should invest a lot more on 5g devices 5g networks, drive this, whether they use different frequency than LTE, how they want to communicate with these devices. I mean, the technology exists, the 5g standards exist. Now the real Applications of how it will help an ordinary human being is not there yet.

- [Ken] Right. Yeah.

- [Muddu] The best example you use Google home or Google home, Alexa, Tesla like this tend to be the still the IoT devices that people are using it, you need a very good Applications of the IoT with 5g that all of us can use.

- [Ken] Yeah. And I think we also need to figure out a sort of personal data anonymization in a way that can be trusted because there's a lot of value in data that absolutely should be kept private. And right now the only answer to that is don't share that data. And so we're not learning anything from it, I think we need to get much better at anonymization in order to make the consumer space, especially really rich enough to to make AI powerful. But we need to make sure that people know that they're going to be protected and and not compromised through sharing their data. I think

- [Muddu] It is. I have a comment on that. If you're into, I can expand to one thing can there is, I don't know, it is definitely do able in the IT industry and that's what we are doing it. I call it the privacy to you. Just send them the, the data belongs to you at home. Ideally what you want the Alexa and Google home to do is can we define your privacy policy?

- [Ken] Yeah I agree. And, and that sort of brings us around as we get near the end of our time. That brings us round to what I think is maybe the most important part of this discussion, which is for the listeners who are often trying to think of ways to get ahead of AI or leverage AI in their business what should they be thinking about? Should they still be in data collection mode? Should they be trying to partner with someone? Should they be deriving their own algorithms? You know, is it simpler to just get involved with a platform company? Like what, what would you recommend for folks who want to leverage AI in their business but don't really know how to get started?

- [Muddu] Very good question. I think the first is I've been debating is through my as I talked to various CEOs and customers, right? Customers, you really need an executor who believes this see what's happening in the industries. Everybody wants to know AI and the ones who put their fingers wet kind of thing but they're really not doing it justice to be persistent. And you got to stay at a project. See there's some projects like this. You want an executive would will next saying that this is my next five, 10 year projects. Most people what they do is I'm going to give a limited budget. I'll do a pilot, I'll do a POC, I'll talk to this. So, and then they claim they're doing AI. So the issue is, first of all you have executives who don't believe in AI should not be fighting. You need somebody who says, look, you believe in the IoT you knew in the podcast, or you bet your career on it. You need people like you who are willing to say, look I believe in this and give them the chart on AI's. That is the biggest problem right now in industrial AI. People are people who are actually running the projects. They should not be running it. They don't believe in it. They are not capable of it or not. Do they have the capacity or the brain trust in that? So that's the number one. Number two is, is data. As you said, collect the data. Now data is your new line. All data is leverageable in filling a model. That's number two. Number three is if you are in a business to build your own solution, then go build it. Then create your own data science team. Yeah, Michelin, if you're not yet what are the solution you want in that industry and work with them, I've ended up. So whether sometimes you are deciding to do it yourself fine, but that's additional. Leave it to what the key is that they should make a right up on the phone.

- [Ken] You know, I've, I've been saying for years that the next sort of company that doesn't exist yet that'll be a trillion dollar company in the next 10 years is whoever figures out, how to make a data stock exchange. So that any company that needs a data set can sort of lease it or buy stock in it. And the data sets are all independently owned or owned in shares by whoever created the data. I swear, there's gotta be a way to do this. And whoever figures it out is going to be the next trillionaire

- [Muddu] I do too. And I put our own company. If somebody gives us various data, you buy them. If you mean the data is clean, good data.

- [Ken] Sure.

- [Muddu] Yeah, you're right. The data marketplace will happen.

- [Ken] Yeah, I think so. And I think the secret to it.

- [Muddu] Is your data. And if they take your data you should get a cut, like the app store app. So if you are willing to share your data you should make money off that. So that tickled on economies should be as Google and Amazon should pay you 30% of the data that they share on you.

- [Ken] Yeah, I agree. I think that that's going to be the secret to these real like people talk about big data sets but I don't think they were there yet at all. To get there, I think the secret is people are going to have to get some sort of investment in or value in sharing that data. And not just individuals, companies you need, you know you need a company like Salesforce to trust that they can anonymize their data and put it into the public marketplace and that they're going to get revenue out of it. If you can get, if you can find that, man, I think you're I think you're pretty much done.

- [Muddu] Definitely.

- [Ken] Muddu, unfortunately. That's about all the time we have for today. Thank you so much for chatting with me today and let's connect. It's been a real pleasure learning about you and about what you've been doing at Aisera.

- [Muddu] Thank you, Ken. Thanks for having me and continue to stay safe.

- [Ken] Thanks again to all of you listening out there. I hope you've enjoyed our discussion and if you have please make sure your like and subscribe. So you don't miss out on any of our episodes we posted every week and I hope you'll leave us a rating review and comment on your favorite podcast platform. If you'd like to be our guest. Please, click the link in the description And we also have a great sister podcast network called the Island. We're all podcasts. So make sure you check that out.

- [Ryan] Hey, Ken, let me jump in real quick and introduce your audience to another awesome show on the IoT for all media network. The show that started all the IoT for all podcasts where I bring on experts for around the world to showcase successful digital transformation across industries. We talk about Applications in IoT solutions available in the market and provide an opportunity for those companies to share a device to help the world better understand and adopt IoT. So if you're out there listening and haven't checked it out be sure to go check out the IoT for all podcast available everywhere.

- [Ken] Thank you, Ryan. Now get back to your show and thank you all for joining us on this episode of Let's Connect. I have been Ken Briodagh Editorial Director of IoT For All and your host. Our music is sneaking on September by Otis McDonald. And this is been a productive media network. Take care of yourself.

- [Narrator] You are listening to IoT For All media network.
Hosted By
IoT For All
IoT For All
IoT For All is creating resources to enable companies of all sizes to leverage IoT. From technical deep-dives, to IoT ecosystem overviews, to evergreen resources, IoT For All is the best place to keep up with what's going on in IoT.
IoT For All is creating resources to enable companies of all sizes to leverage IoT. From technical deep-dives, to IoT ecosystem overviews, to evergreen resources, IoT For All is the best place to keep up with what's going on in IoT.