Analytics Vidhya’s ‘Leading With Data’ is a sequence of interviews the place business leaders share their experiences, profession journeys, attention-grabbing initiatives, and extra. Within the fifth episode of the sequence, we’re joined by a really particular visitor – Mr. Srikanth Valamakanni. He’s the Group CEO, Co-founder, and Vice-Chairman of Fractal Analytics, one of many largest AI firms in India. In this interview, he shares with us his insights and observations about constructing a data-driven group. Being one of many pioneers in information analytics within the nation, he additionally speaks about the altering panorama of AI through the years. Furthermore, his deep ardour for analytics, information science, and training can also be highlighted on this speak with the Founder & CEO of Analytics Vidhya, Mr. Kunal Jain. Listed below are some excerpts from the interview.
The Evolution of AI
Kunal J: I wish to begin together with your very early days. You began Fractal 23 years in the past when analytics was nearly remarkable. So, you’ve seen this business evolve from a really area of interest to the place it’s at present. How has that journey been for you?
Srikanth V: The fascinating factor about our journey is that it has mirrored the journey of progress of AI. AI, as you all know, was a time period coined within the Summer time Convention at Dartmouth in 1956. I had entry to the recordings and notes from the convention that they’d in 2006, which was 50 years after the Dartmouth convention. Among the attendees from 1956, like Marvin Minsky and a bunch of others, had attended the 2006 convention to debate the progress on this planet of AI, in these 50 years. They usually have been attempting to grasp what would occur within the subsequent a few years.
I noticed the proceedings of that dialogue and was fascinated as a result of even in 2006, it seems, individuals have been truly discussing whether or not AI would go the route of first-order logic, by creating guidelines, exceptions, and so on., or in the direction of deep learning, or neural networks, as they have been then referred to as.
Earlier, once I was finding out electrical engineering, the pc science division in my school used to study AI. That they had a course on AI, whereas we had one on neural networks – and so they have been two various things. AI meant guidelines – like fuzzy logic creating guidelines – and we have been doing all these neural community issues like fingerprint recognition, signature verification, and so on., utilizing very fundamental neural networks. This was within the nineties.
So even in 2006, the definitions and distinction was not very clear. And simply 4 years later, abruptly, neural networks and deep studying popped up as core matters world wide. We began seeing spectacular outcomes from the labs of IBM, Microsoft, Google, and others. After which in 2011-12, one thing very attention-grabbing occurred – Google realized that this expertise was going to vary the world in a really important approach. In order that they employed Jeffrey Hinton, who as everyone knows, remodeled Google and added AI to every of their merchandise.
So what I’ve seen through the years when it comes to AI evolution is that there have been the great outdated days. Then by 2010, the largest firms which might be AI natives or digital natives began realizing the potential of AI. And by 2015, the Fortune 100 and Fortune 500 firms from world wide, began to get up to this. Round 2014-15, I began seeing many boards ask their CEOs to come back and current their information technique or AI technique. Nevertheless, it was nonetheless not a really massive deal in India.
However in 2020, it turned actually massive in all places. Particularly after COVID, it began taking off, and now in 2023, after ChatGPT, it has simply develop into a kind of issues that we will’t cease speaking about.
Fractal’s AI Journey
Srikanth V: If you happen to have a look at the Fractal journey, the primary 10 years of Fractal have been all about problem-solving utilizing analytics. So we knew a really clear resolution drawback. So we have been seeing what is occurring and predicting what’s more likely to occur utilizing information, and serving to firms make higher selections with it. Most of this was finished utilizing logistic regression or resolution bushes, random forest, XGBoost, and so on. Till 2010, we have been working with related strategies on structured information for probably the most half.
By 2011-12, we created one thing referred to as Fractal Sciences to discover probably the most spectacular new issues on this area and spend money on them. From this, emerged a bunch of merchandise, and ultimately, we employed Prashant Warrier, who helped us construct Treatment.ai and spawn it from 2015 to 2020.
Equally, we spawned an entire host of AI startups like Crux Intelligence (AI-driven enterprise intelligence), Eugenie.ai (AI for sustainability), Asper.ai (AI for income development administration), and Senseforth.ai (conversational AI for gross sales and customer support).
Between 2015 and 2020, we actually took off when it comes to utilizing AI, ML, and DL to not solely resolve issues for shoppers but in addition create merchandise and companies. And since 2022, we’ve been focussing on constructing our personal basis fashions, our personal diffusion fashions, and utilizing them to unravel every kind of latest issues that we by no means thought we might resolve utilizing information.
Turning In direction of a Information-Pushed Ecosystem
Kunal J: In between all of this, when was your Eureka second; once you personally realized deep studying is actually going to drive this data-driven AI, versus the normal rule and symbol-based AI?
Srikanth V: I’ve at all times believed in neural networks proper from my engineering days. I at all times thought there was big promise right here.
However it was in 2012 or 2013 that I learn a guide referred to as Tremendous Intelligence by Nick Bostrom after which met with Lee Deng who was a really senior scientist at Microsoft. These two incidents made me notice what’s been occurring within the business for the final 3 years and the way a lot I’ve missed out on. In the meantime, even a few of my mates have been publishing papers on unsupervised studying and associated matters. That’s when it hit me that I used to be already late to start out, however I did begin nearly instantly. And that’s once we obtained Prashant, Suraj, and a bunch of others to affix us and construct a crew. Now, in 2023, it appears to be like like we began off about 10 years earlier than most others.
Kunal J: How do you retain your self in control with not solely the enterprise necessities however with the business and analysis aspect as properly? Do you’ve any recommendation for people who find themselves main these information science groups, when it comes to learn how to stability the 2?
Srikanth V: I feel my recommendation to everybody within the business is to get deeper – do your analysis and hold upskilling. Breadth is getting automated. We’re shifting in the direction of a time the place even in case you’re good at shopper administration, problem-solving, machine studying, individuals administration, and a number of different issues, and are very priceless at present, you possibly can develop into irrelevant tomorrow. It occurred within the IT business, it occurred within the tech business; it might occur within the AI business as properly. Your relevance now could be a operate of how a lot you’ve learn or advanced within the final week, not simply the final 10 years.
So my recommendation is in case you don’t have the depth and the sting, it’d be onerous to outlive. So anybody from junior to center to senior administration, I might say don’t lose the tech edge. Spend a while on daily basis to learn analysis papers and keep abreast of this tech as a result of lots is altering quickly. For the junior to media degree, I might say spend a ton of time studying the analysis papers and attempting to duplicate these outcomes. For the senior individuals, If you happen to don’t have that a lot time, a minimum of work with any individual throughout the crew attempting to do the identical. Additionally have interaction with the scientific group, researchers, and school and discover out what’s the following massive factor.
I nonetheless try this. Though it’s tough to seek out time being the Chief of Fractal, I nonetheless handle to remain technically sound and up to date to maintain going and keep related.
Constructing a Information-Pushed Group
Kunal J: How do you see organizations evolving within the subsequent 3-5 years? What would the group of the long run appear to be? How would the work be distributed between people and machines?
AI is probably the most spectacular productiveness development expertise that we now have ever come throughout.
– Srikanth Velamakanni, Group CEO, Co-founder, Fractal Analytics
Srikanth V: Let me begin from the very apparent. Machines and machine intelligence will play a higher position in the way in which we work. It’s going to usher in productiveness development and new intelligence and every part will develop into extra automated. The machine element within the workforce will go up because it occurred throughout mechanization, the Industrial Revolution, the data revolution, and now the AI age. Every of those phases has been about higher automation and the higher position of machines versus people.
The second factor that’s taking place is extra intelligence getting embedded into it. Every part is only a tiny little bit smarter. This progress is compounding progress. So though in a brief span of time, it appears to be like linear, in the long term, say in case you have a look at how smartphones have modified within the final 10 years, the expansion is exponential!
So now, relating to work, I attempt to ask questions, like why ought to we work within the first place? This can be a query that’s price asking. What position does work play in our lives? And whether it is about incomes your each day bread, monetary safety, fundamental wants, shelter, and so on., I hope that the world will get to a spot the place these are assured for each human being. Then what occurs is individuals would wish to work for higher-order wants – like love, ardour, a way of belonging, self-actualization, and so on. As a result of we’ve already created a lot wealth on this planet, it now not is smart to have individuals work solely to earn their each day bread. So in about 10 – 15 years from now, I imagine, we’ll find yourself working as a result of we wish to and never as a result of we now have to.
Additionally by then, the organizational construction would change. I really feel we’re nonetheless within the overhang of the economic period the place individuals are anticipated to work from 9 AM to five PM. Their out and in instances are recorded and their each transfer is being watched. This can change from a time-based system to a knowledge-based system sooner or later. Firms would resemble an artist’s studio somewhat than a manufacturing unit.
Fractal’s Downside Fixing Strategy
Kunal J: One of many key issues that stands out concerning the problem-solving strategy at Fractal is the mixture of AI, engineering, and design. I’ve additionally observed the main target you placed on behavioral design and behavioral economics. So, when did you begin shifting in the direction of that strategy?
Srikanth V: This transition occurred about 7-8 years in the past at Fractal once I checked out the place we have been profitable and the place we weren’t profitable.
We analyzed that and realized two issues have been lacking. One was that we constructed one thing and it took many many months to implement that as we depended externally for the engineering energy. Whereas working with massive firms like Tata and Airtel, though we’d construct the algorithms, it will take them a 12 months or extra to implement it into their techniques, because of this, and we weren’t doing any of that engineering half.
The opposite factor we discovered was by no means to take the shopper’s drawback as a right. They may come to us with a downstream drawback, let’s say ‘not sufficient bank cards are being bought’ or ‘clients aren’t utilizing a particular characteristic as anticipated’ – however we should go upstream from there and path it again to its root trigger, which is the precise drawback to unravel. That is one thing that we discovered very early on – to reframe the shopper’s drawback and work out what precisely to unravel.
So yeah, that’s how we obtained to the recipe of nice drawback fixing that requires AI, engineering, and design, and likewise the artwork of reframing the issue by working with the customers. So round 6-7 years in the past, we introduced collectively a bunch of individuals from completely different disciplines starting from sociology and anthropology to neuroscience and Java coding, and constructed a crew for engineering, the place they might all work collectively and create magic!
Srikanth’s Academic Initiatives
Kunal J: One other side that fascinates me is your ardour for training. Even with such a busy schedule, you’re taking day out to show programs. You co-founded Plaksha College, and you then take a course yearly going there. You additionally train individuals internally at Fractal. What drives you to take that point out? And might you share a few of these efforts?
Srikanth V: Firstly, Kunal, I’m very grateful to my academics. Lecturers have an infinite multiplier impact. You’ll be able to create lots of societal good in case you enhance instructing. If you happen to take a really long-term view of the financial system, training and entrepreneurship are the 2 vectors that may result in better-educated individuals to unravel issues, and entrepreneurs who’re keen to take the danger to unravel these issues. And collectively they will make the world a greater place. Due to this fact, any time that I can spend in constructing that future society or future world, is massively satisfying.
So founding Plaksha College was a kind of issues. However greater than the cash or anything, I imagine it’s the effort and time you set into it that issues. You should submit your self to the trigger. I’ll spend, you realize, 40 hours instructing a course, however the 50 – 100 college students who’ve now gained that approach of problem-solving via this interplay, can go and alter the world. Even considered one of them might change the world in a big approach. And I might take these odds as a result of I feel that via them, we will create a higher influence.
I train a course referred to as ‘Machine Studying to Make Higher Choices’. It’s a captivating course that brings in numerous sciences, from neuroscience and behavioral sciences to problem-solving and cross-organizational processes for making. So the identical matters, once I train the scholars, are slightly extra technical and once I train my Senior Supervisor, are rather less technical.
The hidden advantage of my instructing is that it retains me younger. It retains me on my toes and helps me keep up to date as properly. Instructing is one of the simplest ways of studying. You notice your individual degree of ignorance once you strive instructing a subject. And you then make investments time in studying the subject. You have got a category arising tomorrow you then higher study and also you higher burn the midnight oil in determining that matter. So it pushes me in the proper path as properly.
Kunal J: So as to add to that, you lately launched the Fractal Data Science Professional Certificate on Coursera. Are you able to inform us a bit extra about this information science certificates?
Srikanth V: I feel it’s nonetheless in its early phases. The thought is to create a option to produce nice information scientists in our nation. The business at present has only a few information scientists in India. As of at present, there are 6 million professionals within the IT business, and I really feel like a lot of them have to study AI and Information Science. Furthermore, we’re graduating one million engineers yearly, and they need to know this. So the thought is to make an entire host of those individuals information science literate, with a way more fuller perspective of what it takes to make use of information to unravel issues. The thought is to create a set of extra fuller, extra mature, extra well-rounded information science professionals via this course, who’re prepared to enter the business and resolve these issues.
The Closing Spherical of Fast Fireplace
Kunal J: As we’re heading in the direction of the top of our session, we nonetheless have fairly just a few areas to the touch upon. So let’s simply do a spherical of fast fireplace. What guide did you final learn?
Srikanth V: I attempt to learn as many as 100 books a 12 months. So I’m at all times studying 20 to 25 books in parallel. But when I needed to choose one guide to suggest, it will be a guide referred to as ‘Affect: The Psychology of Persuasion,’ by Robert Cialdini. It’s in all probability top-of-the-line books I’ve learn on any matter. One other guide I at all times suggest is ‘How Will You Measure Your Life,’ by Clayton Christensen. After which there’s Viktor Frankl’s ‘Man’s Seek for That means’. The guide that I just lately learn, which I actually like, is a guide referred to as ‘The Molecule of Extra,’ by Daniel Z. Lieberman.
Kunal J: If you happen to have been beginning out at present, what sort of startup would you construct? What area wouldn’t it be?
Srikanth V: Nice query. I want I knew the reply. I’ll definitely not construct a Fractal-like firm. What I might in all probability do is take a really deep drawback that isn’t solved, work backward to unravel it, after which attempt to construct an awesome firm round that.
I feel constructing one nice firm in life itself is a good fortune, and that too to the extent that we now have spawned off a bunch of startups, which even have their very own path to greatness like Treatment.ai, Crux Intelligence, or Senseforth.ai. These are all excellent concepts that would all develop into nice firms in their very own proper. I feel I already really feel like I’m doing sufficient. So I’m definitely not pondering of a brand new concept alone. But when in any respect I needed to, it will be an concept I resolve one drawback, very, very properly, after which construct a worldwide firm.
These have been the highlights from our unique interview with the Group CEO and Co-founder of Fractal Analytics, Mr. Srikanth Velamakanni. You’ll be able to watch the complete interview here. Keep tuned to our ‘Leading With Data’ sequence on the Analytics Vidhya Community Platform for extra unique interviews.