Ajay Kumar Join us as we have a conversation with Ajay Kumar, Senior Manager for Solutions Engineering at MuleSoft, where we discuss his educational journey, how he transitioned into the MuleSoft and Salesforce ecosystem, and the vital role of continuous learning in technology. 

We also explore the captivating world of Artificial Intelligence (AI). We discuss its evolution, current state, and the promising future it holds. Listen in as we share insights into generative AI, its increasing importance across industries, and its potential to enhance human IQ. 

Additionally, we discuss the significant impact of AI on our daily lives and how it’s revolutionizing industries, especially retail, through optimizing inventory management. Overcoming fear and hesitation towards AI, the necessity of staying relevant in the rapidly changing tech industry, and the irreplaceable human touch in AI are some of the captivating topics we touch upon.

Show Highlights:

  • The evolution and impact of Artificial Intelligence (AI), including its transformative potential across various industries and its capacity to enhance human IQ.
  • Kumar delves into the concept of gendered AI, the current state of AI, and the future potential of artificial general intelligence.
  • The use of AI in the retail industry. 
  • The role of MuleSoft in data integration.
  • The concept of having a personal AI ‘co-pilot’ in the future.

Links:

Episode Transcripts

Ajay Kumar:
I was able to navigate through some of the menus and I could then find out, hey, I can actually use this to add numbers. I can actually use this to generate a square root of a very, very complex number. So, then I learned that to be able to do more, you need to understand something called programming. And what programming meant was you need to be able to talk to a computer, and the only way to talk to a computer is to be able to tell it in the language that it understands.

Julián Duque:
And that’s Ajay Kumar, Senior Manager for Solutions Engineering at MuleSoft. And I’m Julian Duque, your host for the Salesforce Developer podcast. And here in the podcast, we share stories and insights from developers, for developers. Today, we are going to talk with Ajay about data, AI and MuleSoft, but before we will start, just as we lift off, and we often do, with [inaudible 00:01:02].

Ajay Kumar:
This goes way back to the basic days, probably when it was a monochrome, early school days when you had to write the computer program to add two numbers. This is probably preschool days. And I was very privileged my parents having sent me to such a school which had access to one such computer in the library. I used to be that kid who always took that computer’s time.

Julián Duque:
Of course. And you went straight directly to writing code?

Ajay Kumar:
No, this was started by actually figuring out what can I actually do with that machine, because it was a black and white screen. And it was very interesting to see that you actually had to navigate that computer through the menu. This was probably Windows OS back in the day, which was not even the GUI based. It was a CLI based. And those were the ones where I had to enter a few set of commands. And thankfully for some basic computer education, I was able to navigate through some of the menus and I could then find out, hey, I can actually use this to add numbers. I can actually use this to generate a square root of a very, very complex number.
So, then I learned that to be able to do more, you need to understand something called programming. And what programming meant was you need to be able to talk to a computer, and the only way to talk to a computer is to be able to tell it in the language that it understands.

Julián Duque:
Back in those days, because I also remember that the concept of talking to a computer was very foreign to me because sure, how do you talk to a computer? You’ll need to have a special language, a special set of instructions, what are going to be those instructions? So, that for me was fascinating. What was that initial process of learning how to start talking to a computer?

Ajay Kumar:
Well, interesting question. Again, to be honest, I was more of an introvert when I actually grew up. Some of them think, my wife, Padma, she still thinks I am one. And I found it much easier to talk to computers than talking to people.

Julián Duque:
Wow.

Ajay Kumar:
That is probably what drew a lot of fascination in me that, hey, I could tell it anything the way I wanted to without actually being judged or getting any feedback. The only feedback I could get is error, bad code, enter, some error code being thrown at me, but hey, I understood what it means. So, that was how the journey started for me.

Julián Duque:
Oh, beautiful. And what was your first programming language that you used back in the day?

Ajay Kumar:
Basic.

Julián Duque:
Basic. Okay.

Ajay Kumar:
Basic. Yeah. And then I moved to Pascal. I moved to COBOL, I picked up C, C++. And yeah, that’s been the journey before I evolved into Java and the more modern languages.

Julián Duque:
Quite a journey. After that, you decided that this was going to be your thing, you decided you were going to study maybe computer science or something related or how was your educational process after this first experience?

Ajay Kumar:
Yeah, well experience. I really didn’t plan on becoming a computer engineer because I had a very, very variety or wide set of interests. I think I found computers as the easier way to get things done. Back then, that’s probably how my entry point was, computers. And I never thought I’d actually go out and get myself a computer degree, but then it became a part of a necessity at point in time. I come from India where typically the larger career professions for those going into graduation is either you’ll be an engineer or you be a medical professional doctor.
And largely at that time, the time when I was actually picking up my engineering subject to specialize in, computer science was definitely on the top. And I said, “Hey, this can be easy. I know how to do this.” So, I took that up. Well, I’ve always been interested about tinkering systems, on breaking things down and putting them back together. Mechanical engineering would have been even an ideal fit. But then I thought, okay, maybe I want to take the one that is more easier to do so that I can have all the time to do something else.

Julián Duque:
That’s a very interesting approach. Very smart. And that was the reality? Did you have time to do some more things because this was easy for you?

Ajay Kumar:
Yeah, it was actually. And I think it basically gave me different projects that I actually could work on, the time to experiment different things, do a bit of traveling really. And I had these very weird set of hobbies early on, which I haven’t really picked up on. Back then, it was Philately and numismatics. For those who don’t know, it’s collection of stamps and coins. And I have a very, very vast collection of both of them, which I haven’t gone back to.

Julián Duque:
You still have that collection?

Ajay Kumar:
I do. I do. Yeah. It’s somewhere preserved, safely.

Julián Duque:
Well, you need to do some more traveling to grow out that collection.

Ajay Kumar:
Right.

Julián Duque:
Nice. That’s beautiful. And how you got into the Mulesoft or Salesforce ecosystem?

Ajay Kumar:
Sure. MuleSoft has been a more recent journey for me, just from a context perspective. I’ve been now three and a half years at MuleSoft. First started off as an IC, as a solutions engineer, solutions architect as you see it from an industry perspective. And then I actually moved into a leadership role where I currently lead a team of solution architects, SEs for a wider set of industries. And prior to that, it’s typically largely been my journey with the marketing cloud side of the product in Salesforce and MuleSoft was very interesting for me because of how it solves certain set of very interesting problems. I was doing extremely well in my marketing cloud journey at Salesforce to a point where I felt it was, again, going back to my reference of how I felt, it was very easy to talk to computers. It was just like that.
It was in the back of my hand. I could do this half asleep, like marketing cloud. I wanted to pick up a new challenge, and MuleSoft was that interesting challenge and it didn’t disappoint. So, there’s a bit of an interesting story there. I went to Dreamforce 2019, this is a Salesforce’s annual event. It was one of those opportunities to actually be able to pick a booth duty, which everybody as an employee have to do. I volunteered to do a booth duty from Mulesoft, a technology that I didn’t know anything about. I have to be honest here. I knew zero, absolutely zero about MuleSoft, but I’m being one of those personal background that I can actually pick up any skill and learn how to be an expert at it over a period of time. I knew this was going to be one of those areas that I really wanted to pick up. And that was my introduction to at Dreamforce 2019, MuleSoft.

Julián Duque:
And that was your first Dreamforce?

Ajay Kumar:
That was my first Dreamforce as well. I was presenting on stage for marketing cloud.

Julián Duque:
Mine too, so we have a very similar starting year. And I can say that I share a similar way of learning as well. This is how I learned Java or the multiple programming language I have learned in my life is by, okay, I don’t know anything about this and there is the opportunity to build this thing or start this project. I was the first person to raise my hand like, “Sure, I will build it,” even though I don’t know anything about it, I will figure it out. And it was always successful journey for me. I think having a challenge, having something that you can later see a result is a strong motivator to learn something.

Ajay Kumar:
I agree. I totally agree with that. And I think it’s probably one of those skills that many people have different ways of learning, and I really have probably adopted this way of learning over a period of time, and I keep enjoying experimenting with this. It’s like sharpening your ax, or as they call it in the Muley word, if you don’t keep sharpening your ax and you keep using it, over a period of time, it becomes blunt. So, you have to sharpen it again. And over a period of time you have to go and look for newer challenges, newer things to do. That is how it works in my world.

Julián Duque:
Which brings me to a way more modern topic, and it’s something that I got deeply into it a couple of months ago, and I know you might be considered also quite an expert in the area, which is AI.

Ajay Kumar:
Right.

Julián Duque:
Let’s talk about that. We started with a very interesting concept of talking to a computer through a set of instructions, very defined, very specific instructions, which are programming languages. But today you can talk to a computer using your natural language and it will respond in that natural language. I was testing yesterday the new feature, the new voice feature of the ChatGPT voice mobile app, and I got surprised by it. It was, wow, this is something way different that back in the day, it was only possible on science fiction.

Ajay Kumar:
Right, right. I think the possibility is so limitless or endless right now, and you raise a very interesting topic. So, I’m very happy that we got to this topic this soon because it’s one of those topics that’s probably top of mind for everybody around the globe, even if they’re not in the technical field, they’ve been hearing this front and center, even in their day-to-day. How does this impact? Yeah, I’d like to pick some threads on it. My journey into AI particularly started, I would probably correlate this back to some of my work that I did at a university. We did write final year project that was around how do you actually model a programming language to be displayed visually and have some intelligence where we could turn it around and feed a visual version of that whiteboard into that program and can it actually spit out code?
That was something that we attempted very, very crude way to do it, and that was one of my earliest attempts to dip my hands into AI. But now if we dial up a few years later into where we are standing now, the generative AI wave is completely a different game right now. I think the industry is rip, and ready, I would say, to take on this challenge. It probably wouldn’t have been ready to this level of wavelength a few years ago, maybe five, six years, maybe 10 years ago, purely because of the kind of infrastructure requirement and the kind of talent pool that we needed.
And also the kind of data even. I don’t even want to miss talking about data because to get to the right AI, you need to have the right data strategy. And I think that is particularly something that people often forget that you can’t have the right AI output if you don’t have the right data in the first place.

Julián Duque:
And AI is not something new. I mean, we have been hearing way more about it lately, but this is not something new. The thing is that before, you will have to know a set of skills to be able to train a model or work with machine learning. It was not as accessible as it is today with these large language models that are available with generative AI and how everybody’s able to use it, even though they don’t need to have a strong engineering or mathematical skill to be doing AI. So, do you think this is becoming something that everybody needs to start learning or including in their day-to-day jobs?

Ajay Kumar:
For sure. Yeah, absolutely. It’s basically like if you go back 20, 25 years ago, how we started off our journey with access to internet. Very few people out of the university college dorms of Stanford, Harvard and a few other places actually had access to this internet because their systems were wired to each other. They were connected with the university campus network, et cetera, and a few other research laboratories where they were actually dumping a lot of this information. And that was a birth of an early internet.
As we evolved from there, being able to access those information, which was very unstructured back then, and it’s still unstructured to an extent now. I think if you take it to the context that you just raised, it’s particularly coming down to how people will fundamentally use this. And I’ll come back to this topic of why is it largely still going back to that text-based response of how we interact with the generative AI interface right now. But to the point which you asked, everybody should actually be utilizing this. And there’s a human fundamental way to look at it.
A typical average human IQ, and this is not me quoting it, but one of the famous medical professionals quoting it, the average human IQ for somebody who is working, let’s say in a software engineering profession. Let me start with that, because we are all in the software space right now, both you and I, but our audience could be in different domains, different industries. So, the average human working in an IT-related setup is having an IQ of between 100 to 120, 130 could be the max. The moment somebody hits 140, they’re super genius and they start going into a more deeper research and things like that. Einstein, as they said, was having an IQ of about 160.

Julián Duque:
Wow.

Ajay Kumar:
What you can think of, what does ChatGPT give you today? It is basically giving you access to a very, very specific set of information that you can actually query too, much unlike the search engine that you can actually put in a query and gives you a ton of hyperlinks. This gives you a very, very specific format of an answer. What it does today to a normal human being like you and I, is help become that partner who can help boost up some part of our IQ. So, does it actually continue to coexist with us? For sure. I think this journey will evolve into something else.
If people on the call recollect, a recent movie, maybe not so many years ago, there’s a movie called her H-E-R. A very, very well documented articulation of what artificial general intelligence would look like as somebody that could coexist with a human. Everybody will have a system that could exist with them, that could help partner with them on various mundane tasks of day-to-day, but also give them a boost of doing things at a much, much rapid pace. Things that would take you, maybe hours could be achieved in a few minutes, for example.
I think that is where fundamentally it’ll start making a difference. Is generative AI, maybe if I talk about ChatGPT, ChatGPT12, ChatGPT15, is it going to be Einstein IQ level? I don’t know. Maybe it’s a good question to wait and watch. So, what would that level of IQ mapping, can we even map it to an IQ?

Julián Duque:
With the pace of how these technologies evolving, how many years do you think it will take to have properly artificial general intelligence?

Ajay Kumar:
I think because it’s not one vendor or two vendors who are particularly doing this, and the moment the pace picked up on the open source community coming up to do things at an amazing pace, and because there’s so much gasoline poured on this right now with how… It started with large language models, but you’ve typically seen the industry now shifting a requirement saying, “The models doesn’t necessarily have to be large. We can even do with smaller models to begin with, and we can attain the model that we need to size and then get our output done, and work done, et cetera.”
So, artificial general intelligence is probably a few years away. I don’t know, and I can’t put a finger on it yet, but I think we will approach it in phases more than saying that, “Hey, we’ve arrived.” I don’t think that day will come. It’s almost like how we suddenly one day all started getting access to internet. It was very, very inaccessible to a lot of people. Only very few people with access to technology could find it easy to access internet because the natural language mode of accessing internet was made primarily fundamentally shifted after the wider adoption of technology.
So, artificial general intelligence will probably come through phases and we will have version one maybe that’ll start looking like, “Hey, can I dump away some of my old thoughts and notes into it? And I can use it as a recollection center.” This is exactly what somebody like Wolfram is actually doing. Wolfram from Wolfram Alpha, he’s been actually writing notes after notes in his daily, he writes his diary apparently, and he’s been dumping all his diary from the last 30, 45 years into a model, and he’s using that model to query his thoughts from the last 30, 40 years.

Julián Duque:
Wow, that’s very clever. I love all the possibilities of properly using this technology to solve problems. And this is where data comes into play. I mean, we definitely consider this very a data point. So, he’s feeding a model with a lot of data points. Right now, today, if I want to have something similar with my data, what do I need to do? What do I have to do to be able to have my own model or a way to also query or ask questions to my data?

Ajay Kumar:
Yeah, yeah. No, you’re not the first person asking this question to be honest. I speak with a lot of our customers, a lot of our ecosystem partners and developers out there, and this format of the question comes in one shape or the other. Simple answer, to get to the right data strategy, you need to first figure out what are the different sources the data is coming from even? Like you need to plan, you might be a small company, two person company, 10 person company, or even a large enterprise, 10,000 plus 50,000 plus employees, 1,000,001 lakh plus employee size company. And it all comes down to the different sources of data that you would bring the data into the model or even establish a model.
And to find that right strategy to bring data, you first need to plan of an action. What would be that integration strategy in the first place? If you start putting concentric circles one after the other, the first circle for me particularly where I stand, and I always talk about this with my CIO, CXO, CTO audience is have you thought about integration strategy before you even go to the data strategy? Because that’s where some technologies like MuleSoft really bring their strength. We unlock data systems really powerful way, and we help customers realize the full value and potential of that data set. And once that data is available, you can actually understand how much of it is structured versus unstructured.
And on top of that, you can actually realize, maybe I don’t need this portion of data to actually make sense out of my model. A classic example is I told you about the Wolfram Alpha example. Wolfram dumps his day-to-day notes from his diary. But a lot of it is about, “I talked to this person, I talked to this person,” but the most interesting part of his diary, and I had an opportunity to review some of his work when I was in my university days. That’s where he publishes a lot of his diary online.
Some of his very interesting parts of his diary is about the ideas, “Oh, I had this thought today, afternoon while I was sipping tea. Maybe I should write it down.” And that’s an idea that he wants to go back to. And those are the interesting parts of the data. So, as you do map your data, you need to identify which part of that data is really useful versus not useful, and that would really help classify how the artificial intelligence gets used further on.

Julián Duque:
Interesting. You mentioned MuleSoft as one of maybe key pieces for integrating different systems and bringing the right data into play. What about Data Cloud?

Ajay Kumar:
Very good question. Data Cloud is probably the most recent amazing innovation that the company has done in the recent years. I’m very excited about Data Cloud primarily because of the reason that it is natively a hundred percent [inaudible 00:24:19] in-grown from the company. And it’s not something that we’ve thought of acquiring as a product from outside. While there have been multiple acquisitions in the past, but it’s one of those products that is really going to shape and redefine how Salesforce positions data strategy with our customers.
In this wavelength of where generative AI is super hot right now, we are actually setting the foundation to tell our customers that to get to the right AI principles, you need to have your data strategy set in function in first place. And Data Cloud really helps customers bring data not just from the Salesforce ecosystem, but external data. And we’ve made it easier for them to bring data through various channels. There’s the MuleSoft playing a very important role as well to bring that data into Data Cloud, but the real benefit is where customers can activate that data back in terms of insights and actions.
I’ll take one of the examples that probably I’ve used for an exciting innovation use case that we drove for one of the customers, and it’s just when I say me, I’ve also got my team supporting me in shaping this up. It was from one of the largest retail brands here in the UK and sort of use Data Cloud to position and make them understand how, as products get lifted off the shelf, as customers purchase products from the shelf, we need to start think of replenishing the stock. With the Data Cloud, there’s a portion of a segment that gets created wherever the customer actually keeps buying. So, we are streaming real-time data as the products get sold from the store.
And based on the insights captured, we are triggering off a flow to be sent to the logistics management system to actually start sending a shipment of all those products to refill the inventory. And imagine the time it would save and the cost it would save to plan all of this and to be able to do this near real time. That’s the [inaudible 00:26:28]-

Julián Duque:
Totally.

Ajay Kumar:
Yeah, so this is one of the many examples, and I really feel we could do a lot more in this space. And on top of this is where you can see the value benefit of AI doing recommendations, recommending, “Hey, don’t do this, do this better, and optimize this approach.” [inaudible 00:26:47].

Julián Duque:
Yeah, I was getting into that with that specific example, how AI can make things better or more accurate.

Ajay Kumar:
Absolutely. Absolutely. And I think it’s definitely very, very convenient now. You saw how we’ve now got vision support within ChatGPT. Many of them have started uploading images and they’re asking it to translate. Imagine the power of being able to read off imagery data, for example, in this previous use case I mentioned from a shelf or what’s happening in the store from a store map at a very high level and being able to translate that to how can you optimize your store’s inventory? Where do you place the products best so that customers can actually pick it off faster? All of those combinations, instead of having to do experiments after experiments, you come out with the best possible combination. Those could be the kind of smart recommendations coming up of the many other scenarios.

Julián Duque:
Indeed, a fascinating new world that we are living right now, but still there are a lot of people that are hesitant to start using AI for multiple reasons. Do you have any advice for our audience or folks that still, they are maybe scared of it or they don’t want to use it or have some sort of fear of AI taking over our jobs or a lot of different aspects of our lives? What’s your thought about that?

Ajay Kumar:
Yeah, I think the analogy that I can recollect or relate to this is imagine those times when I know we haven’t fully gotten away from those days. Imagine those times when the first time a digital money was introduced to the world, people were hesitant to say, “Hey, what is this money? I don’t actually see my money anymore. It’s somewhere in the format of a record. I actually don’t touch that physical cash anymore.” We still do some cash transactions, cash in hand, but almost 95% of the transactions globally today are all digital. The 5% is the ones that are typically hand-in-hand cash transactions.
And that is typically the place where we need to get to. The early adoption would be for enthusiasts like tech enthusiasts who have been fascinated about getting on board early. And I think we have those of us who have been interested. But I think where a larger part of the goal that is missing is the piece that lies with the missing link with education and the wider education system in terms of enablement and spreading awareness of why this is important and how this will make their lives better. I think that’s the approach we should be taking. And coming back to the job security aspect, it’ll actually make them even more skilled and highly efficient.
Probably, it’s going to help them earn more money as their skill set improves because it’s exactly how their livelihood could be improved. So, I have three recommendations if I could give for everybody.

Julián Duque:
Please, please go ahead.

Ajay Kumar:
Across different age categories. This is what I told my parents the other day, and this is what I would also tell my daughter. She’s two now, Abha, two years old, but as she grows up to be maybe four or five, I would tell her the exact same thing, live outside your comfort zone. The comfort zone for us today is, “Oh, I don’t really need to do that. The AI stuff, I don’t need it. I can live without it. It’s okay.” That’s live outside your comfort zone. Try to get outside of your comfort zone. Reinforce the human touch of AI, because I think where people fear most is, is AI going to take over my job? Is AI going to be able to do everything that I am able to do and then I won’t be needed? My requirement wouldn’t be there in the company.
For example, in our role, my team does a lot of pre-sales solution engineering demos, and they build a lot of POCs and they help evaluate technology for our customers. Imagine AI taking over control of a lot of that. So, reinforce the human touch. There will always be the human in the loop, which is not going to go away and constantly think about how you need to stay relevant to the change in the industry. If you want to be shaping up that ax, sharpening that ax, I think that’s the one that you should relate this to. And no matter which background you come from, either you’re in tech or you’re not in tech AI, one way or the other will continue to have an impact in your day to day.
I was talking to a friend the other day who told me about, he’s in the medical profession. He told me about how they’re now be able to use the radiology scans to be able to detect some of the diseases that could possibly be shaping up in very, very early stages. And they’re feeding off data from multiple radiology images that have been collected over the world, over these many years. Of course, preserving the confidential data of the patient, of course, not to revealing who that patient was. They’re able to actually model and say, “Hey, looks like you need to be on this medicine because in two years time you’ll start developing an early stage of cancer, et cetera.” So, things like that could really help.

Julián Duque:
Yeah. That’s very, very powerful in there. Indeed, this is a fascinating world. And now that you mentioned your daughter, right now we have a lot of what they call digital natives. Now we will start having AI natives? That might be a term maybe.

Ajay Kumar:
Maybe. Yeah. I think by the time I see my daughter ready to go to school, I find she would have a co-pilot with her who would help her with her questions. And I think it becomes a part of our education system as well where your assignments, your homeworks, all of it could be vetted. It could be the second pair of eyes, which could help you augment your intelligence much, much faster.
And you could achieve some of those manual human tasks, which you could decide to switch it off to. You can always go back to doing the old way. Not to say that, but as the world is pacing forward that way, I think everybody will continue to have a co-pilot one way or the other. So, that’s where I see my daughter maybe in 10 years time having a co-pilot of her own, which is due to her own model.

Julián Duque:
Interesting. Our own Jarvis, our personal Jarvis. That’s very interesting. Final thoughts, how can I get up to date with everything that is happening? Do you have any resource that you want to share with our audience? And of course, we can always add links to the show notes.

Ajay Kumar:
Yeah, yeah, absolutely. I think there’s enough and more content available out there. From a Salesforce perspective, we are doing amazing job on our Salesforce Trailhead. So, for those who are on the Salesforce ecosystem, please go into Trailhead and if you put in the keywords artificial intelligence, you’ll find amazing set of trails that will give you the basics of artificial intelligence to get started. And there’s both the predictive as well as the generative, but for the wider ecosystem who are not on Salesforce, you should definitely tap into all the free courses that are available out there.
It could be the Courseras of the world, it could be the Udemys, or it could be even what likes of Andrew Ng is doing so very well with his deeplearning.ai portals. There’s a lot of free courses, which start with the fundamentals, start going deeper and deeper depending on the preference of choice of topic that you want to go into. As you said, we’re happy to share those links here and there’s a lot more other free courses available. A lot of universities have now started offering, Harvard, Stanford. They want people to learn and pick up these courses without actually paying a dollar price.

Julián Duque:
Yeah, the last one you mentioned, the deeplearning.ai is the one that I’m doing right now and I have learned a lot.

Ajay Kumar:
Amazing.

Julián Duque:
Is an incredible resource. And of course I’ve done most of the different modules on Trailhead. A lot of great information in there. We will be adding the links in the show notes. Ajay, thank you very much for being here with us, sharing your insights about data, AI and MuleSoft.

Ajay Kumar:
No, thank you.

Julián Duque:
I’m looking forward to have you here in the future, of course.

Ajay Kumar:
Absolutely. It’s been such a pleasure to be able to speak about the three favorite topics of mine, and I really hope people find value in this and always reach back again through the network of this link. And thank you once again for having me.

Julián Duque:
And that said, if you want to learn more about this show, head on to developer.salesforce.com/podcast where you can hear all the episodes and read the show notes. Thank you everybody, and talk to you the next time.

 Get notified of new episodes with the new Salesforce Developers Slack app.