More Intelligent Tomorrow: a DataRobot Podcast

Applied AI - Innovation That Matters - Debanjan Saha

June 09, 2022 DataRobot Season 2 Episode 17
More Intelligent Tomorrow: a DataRobot Podcast
Applied AI - Innovation That Matters - Debanjan Saha
Show Notes Transcript

In this episode of More Intelligent Tomorrow, we hear from Debanjan Saha, the President and Chief Operating Officer for DataRobot. He sat down with host Ben Taylor to discuss how he got to where he is and what he thinks is in store for artificial intelligence and machine learning.

Ben starts off by asking Debanjan about the unique journey that brought him to the world of AI/ML.

While Debanjan had originally planned to be a professor, a summer job at IBM Research started him on a climb up the technology stack. From systems and networking at IBM and a couple of start-ups, to databases and data lakes work in Amazon, and analytics with Google, the next logical step was to take on AI/ML. That brought him to DataRobot.

He says IBM Research was filled with talented technologists, Amazon was all about executing growth at scale, and Google taught him about building innovative distributed systems. He learned that experience is a key skill, and it takes time.

"There's no compression algorithm for experience." - Andy Jassy

Every technology goes through a hype cycle, where it’s the “next big thing.” It was the internet in the 80s, the dotcom boom in the 90s, and big data at the start of the new century. This decade looks to belong to artificial intelligence and machine learning.

The good thing about excitement around innovative technology is that it can lead to more investments, and investments can compress development time. It’s part of the growth phase.

"We will use AI without knowing that we are using AI. It's probably going to be in pretty much every decision-making process that we have.”

For AI/ML to succeed, we need to set proper expectations. AI can help you get to better decision-making, but there must be a return on investment to make it worthwhile.

Another expectation to be set is that it’s not just about technology. It's also about process and company culture. Software alone isn’t going to solve your process problems or your culture problems. 

It will take time and experimentation to see how AI/ML will work best for any given situation. And don’t forget that failure is part of the process. Fail fast, iterate quickly. Find what doesn't work, and move on from there.

In healthcare, Debanjan notes that AI/ML is augmenting what humans can do in symptom analysis, image processing, and remote diagnostics. And in the field of sustainability research, it’s being used to develop sustainable harvesting solutions.

Debanjan goes on to share with Ben some of the most memorable lessons he has learned working in technology.

First is that plans never work. But a close second is to be curious. Do that, and you'll end up in the right spot.

"I think what’s most important for a technologist is to be intellectually curious and continuously learning."

Ben wanted to know how Debanjan's experiences in sales and marketing have influenced his technical decision-making. His advice is to focus on customers and how you can delight them. Know the problem the customer is trying to solve. Also, think big. Don't think 10% bigger, instead think 10x bigger.

Finally, Ben asks what advice Debanjan would give to people starting out.

Imagine where the world will be five years from now and work toward that. Think Big. Take risks. Be curious and learn.

Listen to the full episode to hear Debanjan speak on:

  • His tech journey
  • What he thinks of the hype around artificial intelligence
  • The keys to successful AI/ML integration
  • His thoughts on the future for AI/ML
  • Why he suggests getting advice from outside your circle

Debanjan Saha (00:00):
We will use AI without knowing that we are using AI. It's probably going to be in pretty much every decision making process that we have, every application stack that we have. It's almost like software. Probably, we didn't imagine that 30 years back, but today's software is everywhere. All is driven by software to some extent. And I think AI has the potential to transform every aspects of businesses and lives the same way we have seen with some of these pervasive technologies like software and internet.

Speaker 2 (00:30):
Welcome to More Intelligent Tomorrow, a podcast about our emerging AI driven world, critical conversations about tomorrow's technology today. On today's episode of More Intelligent Tomorrow, Post Ben Taylor sits down with Debanjan Saha, President and COO at DataRobot.

Ben Taylor (00:55):
Debanjan I'm really excited to talk to you today. You have a very unique background. You worked for IBM, Amazon and Google, now DataRobot and I'd love to hear more about your backstory.

Debanjan Saha (01:04):
Ben, same here. Great to see you and great to talk to you. It's a long story. I don't know if you want to get into all the details, but I'll give it the shot. Tell you the truth, I never wanted to work in industry. I wanted to be a professor, my father is a professor, my uncles are all professors. So when I completed my PhD, I primarily interviewed in academia and was almost going to be a professor at University of Michigan at Ann Arbor. It was a last-minute decision that I made. I did a summer job in IBM Research and they offered me a good option. And I thought it would be good to be in industry for a couple of years before I go back to academia and teach. That's how I started. At that time, I used to work in networking and this was a deliberate decision and I thought networking is the right thing to work on at that time, this was early days in the internet.

Debanjan Saha (01:55):
So I was in IBM Research for a few years. After that, I went and did a startup in optical switching. We did an IPO and then after the market didn't cooperate to some of us, I remember during the dot-com bust, I came back to IBM and then went into work on the IBM Systems and Technology Group for a few years and built various different interesting distributed systems and storage. And then got an opportunity to go and work in the IBM lab in Shanghai, China. It was a great experience for me and my young family at that time. I finished that assignment, came back and then I moved to Bay Area to do another startup and got picked up by Amazon to work on very interesting relational database project to call Amazon Aurora.

Debanjan Saha (02:45):
I was not quite sure at that time whether that will be a successful project or I will be successful doing that but I took that opportunity anyways, and that became probably one of the most interesting phase of my life and one of the most probably successful project and product I worked on. Amazon Aurora some of you know was the fastest growing service in the history of AWS. I think it still probably is. I was in AWS for roughly about five years, ran all the relational database business for some time and then moved to Google to do data analytics. As you can see, I climbed up the stack from infrastructure to data, to analytics. And I think the next stage, logically speaking, if you extrapolate, I know a lot of people in AI do their modeling and you can clearly predict where I should land and that's AI/ML. And that's why I'm in DataRobot.

Ben Taylor (03:39):
What were you focusing on for your research that you liked?

Debanjan Saha (03:42):
So I was actually focusing on the traffic management in the internet to various different types of protocols. As you know, for example, TCP, ATM was popular. At some point I did some work in that area. Also I did a lot of work in traffic management in the internet, so that's what my research was on. When I moved to the startup, I was using some of these internet protocols to manage the optical networks, which was not quite the case because they were very hardware-centric. They are not software managed. And I actually worked on building some protocols, borrowing ideas from the internet and applying that to optical network. And that became a major internet standard. I also wrote a book on that and a lot of people assume what I did research and I became an IEEE fellow for my contribution to networking.

Debanjan Saha (04:34):
Then I actually moved to something slightly different. There was a lot of interest in network traffic analysis. And I got into that and did a lot of analytics to analyzing a network traffic that was going through network for various different reasons, for security purposes, for traffic management purposes, et cetera. That was my first foray into analytics coming from networking background and doing network analytics and traffic analytics.

Ben Taylor (05:02):
You've worked for some iconic companies. Many of these companies have written books and they're quite famous for cultures for innovation for the way that they've really pushed the boundary. So what have you taken away working for these companies? What are the lessons learned?

Debanjan Saha (05:17):
I learned different things in different companies. I remember I think the month or probably a few months after I landed in IBM Research, the IBM supercomputer beat Kasparov in the chess game. And that was a fascinating moment for me and for the whole company. In IBM Research, I really enjoyed doing technical work. I learned from a lot of people who used to be in IBM Research at that point. And there was some really cool stuff that was going on there, both in the infrastructure side, as well as the software stack and AI/ML side. Some of this went into IBM Watson many years after that.

Debanjan Saha (05:54):
When I moved to Amazon, I think one thing that I learned from Amazon is how to execute well. And Amazon is an execution machine. I haven't seen anyone execute as good as Amazon through that experience. Other interesting thing I tell people is that I have been in large companies which scale very well. I've been in small companies which move very fast. And I think Amazon is one company which moves very fast while scaling immensely. That's of course is a very difficult problem and unique challenge to handle. Some people don't grow with the scale. And I was learning a lot of things I didn't know, so it's a great experience for me.

Debanjan Saha (06:31):
Google is a slightly different culture. I think of all the company that I worked, I probably learned most about how to build really, really innovative distributed systems at scale in Google. Google is one company which believes in building new software for everything that they do. Most of it is really, really good. And that was a great experience for me, working on products like BigQuery, Dataflow, et cetera, which are iconic products, both internally for Google internal infrastructure as well for a lot of customers who use them.

Ben Taylor (07:04):
One of the themes I'm thinking about listening to you is experience is a key part of this game. So that experience of rapid growth, where you're having to double your team up and higher, I imagine you learn and you have to adapt. I also love this concept of forced innovation from rapid growth.

Debanjan Saha (07:21):
That's definitely true. And as one of my managers used to say that there is no completion algorithm for experience. You go through that and you learn from that.

Ben Taylor (07:35):
So the really fun thing about your history is you've been around long enough to see these stops and starts with AI. So I imagine when Deep Blue beat the world's best chess player, there was probably shocking, eye opening, if AI can do this, what else can it do? And in neural networks, we've talked about two neural network winters where there's been hype. When they first invented the neural network, they said, this will do everything. It'll learn to speak and read. I think this was in the '60s and '70s, and now that feels more believable, but where are we on this hype cycle with AI today? What would you say to an executive that hasn't engaged in AI yet and maybe they are concerned about the hype?

Debanjan Saha (08:18):
I think it is somewhere in between hype and where I believe it's going to land with respect to its potential. Every technology cycle goes through this. I have seen this multiple times and participated in multiple of these cycles, starting with the internet cycle and the dot-com bubble that came with that, then Big Data. I don't think there was a huge hype cycle, but there was definitely somewhat of a hype with Big Data that's going to solve world's problem. And I think AI/ML is no exception to that. In fact, AI/ML has been around for a very long time. I remembered taking many AI/ML courses when I was a grad student many years back. And of course it had solved many interesting problem on the way. I think it's only recently, I think we have seen, there is little bit of I think exuberance with respect to AI/ML being used in pretty much anything, right.

Debanjan Saha (09:13):
And I call it salt and pepper. You can sprinkle it over any food that you have and it'll magically taste better. It does taste better by the way. But I think often time when you go through this cycle, there is probably more excitement than the technology can handle or the value it can produce. There is more investment as a result of that. And when you have more investment and we don't satisfy all the expectation, there is disappointment and ultimately things settle down. I do think if you look at the internet and the dot-com, some of the things we probably anticipated 20 years back, it's happening now, it just took a little bit longer than what people expected. If you look at Big Data, it is driving digitization and digital innovation in many sectors of the business. If you look at AI/ML today, it's being used everywhere.

Debanjan Saha (10:03):
In many ways, it is helping businesses. It's helping societies, probably there is a gap between what we can deliver versus what the expectations are and how quickly we can deliver them, right? What happens in these hype cycles is that there is over investment as what can be done also get compressed. Things that probably would've happened otherwise over 30 year period, it get compressed to 10 years, people expect to be done in six months. We often don't satisfy that heightened expectation, but the hype cycle also helps with respect to investment and driving growth faster than it otherwise would've happened if we followed the normal business cycle.

Debanjan Saha (10:41):
So it's not necessarily a bad thing. I think AI is going through some of that, but I do fundamentally believe that AI has the power to transform every interaction in every business, in every industry, right? Whether it's going to happen at six months from now, probably no, but over a period of time, and this could be five years, 10 years, I think it is going to transform the way the other hype cycles like internet has transformed both businesses and societies.

Ben Taylor (11:09):
That makes me think of the database examples. So it'd be almost impossible for you to find a business that doesn't use a database somewhere. Whether it's point of sale, your barista, they're going to use a database. So do you think AI will get to that point where in the future, we will have a very difficult time identifying a business that isn't leveraging AI?

Debanjan Saha (11:31):
I totally believe in that. It's not only that. I think we will use AI without knowing that we are using AI. It's probably going to be in pretty much every decision making process that we have, every application stack that we have. It's almost like software. Probably, we didn't imagine that 30 years back, but today's software is everywhere. Right? All is driven by software to some extent. And I think AI has the potential to transform every aspects of businesses and lives the same way we have seen with some of these pervasive technologies like software and internet.

Ben Taylor (12:11):
I want to unpack this, the failures and the wins, because we exist in a market today where I think the common statistic that's thrown out, there's 85% of AI projects fail. So there are examples in the market of failures, but there are also examples in the market of transformational wins. And you have access to various customers that are finding huge wins using AI. What is the key difference?

Debanjan Saha (12:34):
I think there are different dimensions to it. Part of it is about I think setting expectations. Sometime people expect too much out of what they can get. Of course, AI can help you with decision making. AI can help you getting to better decisions, but there may or may not be enough business value based on what workflow or process that you're trying to improve, for example, right? Sometime the decisions are very impactful, sometimes they are not, and there is a business correlation between what we are trying to impact and the value that you're getting out of it. In most cases, we get some value, but sometime it takes longer time. Sometime the value is probably 10% improvement versus 10X improvement. Right. And I think we have to set the expectation and that takes some time to settle down.

Debanjan Saha (13:23):
Second, it's not just about technology. It's also about business processes, people and the culture in the company, or the business or wherever we are trying to apply AI. And that also takes some time. Yes, you can have an awesome technology, but a lot of times it's hard to change business processes. You have to embed AI in the right point in the decision making process, the right notes in your workflow, and you also have to bring your organization with it. That requires sometime cultural change, sometime more training education, sometime management intervention. And all of this time, all of these things take time. And we have to be thoughtful, deliberate with respect to setting expectation and also planning it well, not applying some software or buying some software is not going to solve your process problem or your culture problem.

Debanjan Saha (14:22):
You have to train the team, you have to get them excited about what it can bring. And sometime people are justifiably worried about what AI can mean to their jobs. For example, a lot of people, when you have these transformational changes, people are worried about is this going to impact? And those things also have to be handled the right way so that we can bring everybody along. Every time you have this transformational changes, there are societal impact that we also have to think about and making sure that we handle that with finesse and with thought.

Ben Taylor (14:58):
And there's been a history of these massive transformations with technology. You had the Industrial Revolution, you had computers and the internet. And I think with all of these changes, massive changes in society, people have to adapt and often will have jobs that are created, but we will have jobs that go away. So you can imagine someone in the courtroom whose their job is to transcribe what's happening, eventually AI will be to the point where it is superhuman with speech recognition. We don't particularly need that role. So I'm very optimistic because I feel like with some of these changes, we see it actually unlocks creativity. It allows people that we're trapped in a process to join a new process or focus on a new part of the business or invest more in innovation. So maybe I'd love to pick your brain to understand how does AI accelerate innovation in an organization.

Debanjan Saha (15:47):
Well, the way I always think about applying any improvement, any innovation into a particular domain or an organization, thinking about what is the problem that we are trying to solve? Who is on the other end of the problem as our process, as the customer? And start from that customer "point of view" that what is that they're expecting from this and what is that we can improve for that customer? If you take that customer view of the problem and walk backwards from there to figure out what is that we need to do to improve the satisfaction of that customer, delight that customer, it works as a north star for me, figuring out some of the difficult decisions sometime we have to take, right?

Debanjan Saha (16:40):
And sometime there are conflicting decisions. For example, whether something we automate versus provide a service which is more manual, right? There are different implications of automating it from the point of view of person providing that service versus person who is on the other end of that service, who is the customer for that service. But keeping that customer centric view and walking backwards from there to figure out how we can delight that customer, improve that process, bring more accuracy, provide that faster. Is often the right way of thinking about it, I think. And I think if we use that rubric to decide what you need to do in terms of bringing technology, what you need to do to improve your processes, what you need to do to some of this human task that you need to, I think, changes thing in a way that ultimately leads to a better outcome for everyone.

Ben Taylor (17:37):
So it's not about having a black box model that's spitting out predictions. It's about who are the subject matter experts, who knows the process. But sometimes automation is not the answer. Sometimes it's some type of augmentation, where humans in AI are working side by side. I think too often, AI is the shiny new toy. Wouldn't it be cool if we applied AI to this use case? And that's been interesting to me, because it's almost like the executive has forgotten the value.

Debanjan Saha (18:04):
There's always that... I wouldn't call it danger, everybody gets excited with shiny new toys. So only after using it for some time, they understand the real value of it and real potential of it. And I do think you need to do some experimentation before you figure out how much of AI/ML can help you and in what timeframe, in what particular sequence and how do you prioritize some of this. And it's not always possible to have this crystal ball to say that, Hey, AI is going to be more applicable here versus here. And that experimentation is pretty natural in the maturity cycle of adopting a particular new technology.

Debanjan Saha (18:46):
So I don't really have any issues with that. I think what I think we need to focus on is that keeping a data driven, value driven attitude towards, yeah, I don't know how this is going to impact me, but let me try and see what I get out of this. And if I see value, then I double down. If I don't see value, then I step back and see how I can create more value. That is the right way of thinking about it because nobody can guarantee that something is going to be successful until it is actually successful, right? Hindsight is 2020. But until we get to that point, and I don't think we are at that point yet because there's so much of unknown, we have to do some experimentation. We have to take some risks. Some of these guaranteed to be fail and that's part of the maturity process.

Ben Taylor (19:38):
You're bringing up an excellent point and that is, you can't guarantee success for particular projects, because sometimes it comes down to the data. It's outside the control of technology. But the thing that I think we can control, the thing that we can fight for is time to value. And how long does it take us to fail?

Debanjan Saha (19:54):
I totally believe in that. I think I don't know where I learned this. I believe in do new experiments, fail fast, bring product or services to market quickly and iterate quickly. I think that's the best way of getting ahead and creating more value over time. And I used to call this minimum lovable product, not the minimum buyable product, right. That you create something quickly and you know there are rough edges, but get it in front of the user, so the customers. Get their feedback and create discontinuous feedback loop and iterate quickly.

Debanjan Saha (20:37):
And it's not just AI/ML. If you look at cloud, it's also created the environment where people can do big experiments, fail fast and learn from it and move on. But what we need to do in order to support this is to reduce the cost of failure as much as possible. There is always going to be some cost if you fail, but if you can reduce the risk of failure and cost of failure, then you encourage people to take bigger risk and do bigger things and raise the bar higher. And that's how you make progress and take bigger step forward.

Ben Taylor (21:13):
I completely agree. And I think failure is dangerous because if you invest a lot of resources into one project and if it fails 12 months later, you've lost all political capital and momentum to try again. And so I think we do celebrate the crawl, walk, run, get the win, get the value. And then you can take higher risk of bets.

Ben Taylor (21:35):
Now that you've been ramping on AI and... You've been ramping on AI for quite a while. What are the opportunities and challenges in the future that you see from an industry perspective?

Debanjan Saha (21:52):
I think about this in a structured layer way. That's the way I have grown up. So when I look at the opportunity in the AI/ML space, I see this, there are opportunities, different layers of the stack, right? Let me start with probably at the ground level, you have the infrastructure and there's a lot of investment going in, of course, in GPUs and the TPUs as well as various different hardware accelerators and there's lot of investment that's spanning in that area. I think that is a fertile area. And when I was in Google as well in Amazon, we had lot of customers who will just ask us to provide really scalable infrastructure, hardware infrastructure, where they can do their training. They can do their inferencing and writing the software from scratch because that's the way they were. They needed just the infrastructure, very, very scalable infrastructure.

Debanjan Saha (22:45):
And both Amazon and Google, for example, invest a lot in that infrastructure for their own internal usage. The Google search engine, for example, is a huge user of ML infrastructure. If you look at Google's ads, there is lot of modeling going on in predicting behavior and that's all use that infrastructure. And we had lot of customers and I probably shouldn't name them, but they use that infrastructure. And they're only interested in that infrastructure.

Debanjan Saha (23:15):
On top of that, there is a platform layer where these are people who don't want to build everything on their own. They want a platform so that they can train and deploy their own models, but they want supporting infrastructure for that. Right. On top of that, there are another type of customers who want pre-train models and they just want to use these models probably through an API. And this is where, for example, things like Document AI comes into play, right? And there are examples of providers who are providing that pre-trained models, sometime very, very sophisticated models with hundreds of millions of parameters, which they have trained. And there is a big business in that area for specific domains, could be document, could be speech, could be video. And I think there is a big market for that.

Debanjan Saha (24:11):
And on top of that, you have various different AI driven applications for different particles, different use cases. And I think a lot of these are popping up and I expect to see many more of this, right? So these are in my view, the different layers in the stack and different opportunities, depending on what type of customers you're trying to give it to.

Ben Taylor (24:42):
Are there any concerns you see in the future with AI being... we talked earlier about cybersecurity or ethics, sometimes AI has unintended consequences. What are some of the concerns that might be in front of us?

Debanjan Saha (24:57):
Well, anytime you have technology this powerful, there are of course concerns it can be used in ways which are not ethical. And AI is no exception. I think we have to have clear focus on what is the right ethical use of AI and how we make sure that we help people stay within that guard rail. Sometime people do that intentionally. Sometime they do that unintentionally. And when it is intentional, it's probably not always in our hand, but that's something that we should always be careful about.

Ben Taylor (25:34):
And the unintentional that you mentioned it's pretty common in industry. So anytime there's been bad press around some AI use case, most likely it was unintentional. They didn't proactively think about the risk and they were caught off guard. And I think that's something that we've had a good culture internally where we've proactively thought about all of the risk, but the exciting thing... I know this conversation is maybe a little bit more concerning, but the exciting thing is if we nail the trust part of it, if we're very proactive, then that opens up opportunities in healthcare. What are some of the bigger impact AI applications that could change the lives of the future generation?

Debanjan Saha (26:16):
Yeah. There are so many of them, right. I think most interesting one to me are when AI is applied to healthcare, for example. This is definitely an area where AI is not so much artificial, but it's more augmentation of what humans can do in either identifying symptoms proactively or identifying, for example, whether through image processing or other technology... Well, people operating AI can be a huge augmentative tool in helping those surgeons, doing those, doing their day to day task. Right. I think there is a huge potential in doing that.

Debanjan Saha (26:57):
I also think in terms of sustainability and some of this societal aspect of how do we make sure right distribution of resources, in areas like that AI can play a vital role. And I have seen, for example, from geospatial aspects to sustainable harvesting and areas like that where AI has been used already, and there are a lot more potential in those areas where AI can be used. In terms of, for example, fairness in some of the decision making that we make, AI in my view, can be more fair sometime than humans because we always, not always, but we have the tendency of bringing our own biases and all view into decision making and which has unintended consequences. But hopefully AI can be more fair because it's more data driven process.

Ben Taylor (27:56):
I love the mention of bias. Because I think sometimes as we gather experience in life, we hope that we become smarter and we hope that we become more creative, but sometimes experience will bias you and it can throw you into a rut where you'll be more... you'll either pull a decision, whether it's an appraisal or something where you're pulling in your own experience or it will limit you to innovation. So it's interesting that bias is something that is ever present with humans.

Debanjan Saha (28:22):
That's right. And that's right. And with more experience unknowingly, you might introduce more bias.

Ben Taylor (28:27):
Yes. The experience theme is very powerful in the healthcare setting. Because you can imagine when someone goes in the hospital, they're dependent on the physician's experience. So how many surgeries has this physician done or the surgeon? Is it 10? Are they new? Is it 100 or is it 1,000? That as we begin to leverage data, you're now benefiting from not one physician. You're benefiting from all of the data in California, all the data in the US, or eventually all the data in the world to inform what the next spec's best action is, which I think is a... that's a very exciting future to think about because we're not there today.

Debanjan Saha (29:04):
Correct. And it also helps with respect to, for example, providing care, right? You have the world's best physicians probably concentrated around the world's biggest cities. If you have to provide care to a rural area, right? Let's say you are doing cancer diagnostic, you may not the train physicians in every part of the world, but you can for example, take a picture with your cell phone or use some other diagnostic hardware that easier to distribute around the world. And you can collect that data and do remote diagnostic or automatic diagnostics, which will make higher value care or more sophisticated care available, more widely towards various different parts of the world that I have seen some companies do. For example, IBM Watson was doing that, doing cancer diagnostics in the rural part of China, for example, which was pretty amazing, which was not available before that was done.

Ben Taylor (30:05):
That's a very exciting application because building AI systems and exporting them. Not that I intended to have this available, but this is an Nvidia Nano, I think it's a $60 computer as a GPU, but these things just get smaller and smaller. But like you said, an iPhone is incredibly powerful. So having something like a mobile device as the technology is vetted out and trained, it's quite easy to export that all over the world, which is exciting. Because of your career and your experiences, you've worked with a lot of different types of teams, different priorities. And this next question is maybe a little bit more difficult to parse through, but what are the most memorable lessons you've learned working in technology?

Debanjan Saha (30:52):
Well, one thing I have learned is that plans never work and interestingly, whatever I plan for myself in different career turns that I have taken, I plan for one thing, I ended up doing something different, but eventually ended up in a reasonable place. So I think what most important for technologists is to learn and be curious. And if you follow that path, I think you'll ultimately end up at the right spot and pre-planning everything is very, very difficult, right. And I never imagined that I'll be at this place 10 years back, or I never imagined that I'll be in an AI company when I graduated some 20, 25 years back. Right. And what has been constant throughout though is that I have been learning as a programmer, and different dimensions of learning. Sometime it is technology, sometime it is management, sometime it is customer obsession, et cetera. And if you keep your eyes and you're open and open to learning, I think you always end up in a reasonable spot.

Ben Taylor (32:10):
And having that breadth of experience I imagine it informs a lot of the decisions you make. So the fact that you've worked with customers, that you've been involved with different aspects of success and sales and marketing, how does that shape your technical perspective? So when you're making decisions on technology, how does the rest of that influence what decisions should be made?

Debanjan Saha (32:34):
I think there are a couple of things. One, I think focusing on customers and how you can delight them. That has been at least more recently, the clearest north star for me, that what is the customer problem trying to solve and walking backwards from there, coming up with what is the technology that I need? Not the other way around. Right. I think that has helped me quite a bit in terms of picking various different paths and sometime making the right decisions, keeping the customers in the center of whatever I'm trying to decide.

Debanjan Saha (33:08):
The second thing that I think also important is building a good professional network. And that's where I get most of my ideas from. A lot of my ideas actually come from people who not exactly work in the same field, but probably slightly different, sometime a lot different field. Right. Of course, I talk to a lot of customers. I talk to lot of startup founders. That's one thing I enjoy doing, sitting in Silicon valley and spending my weekend in cafes, talking to various different founders. I talk to a lot of VCs, for example, see where they're seeing business moving and where they're investing, and cultivating all of this to figure out where the park is moving has been very, very fascinating for me. And it has helped me guiding both my own career decisions as well as figuring out where we need to invest, for example, is a part of the business, or where the opportunities are, or what the next turn on the crank in a technol.ogy trend is going to be that I think is very, very important also.

Debanjan Saha (34:17):
Last thing I would say that I do like to think big, the two things I think is important for decision maker or technology executive, one is looking around the corner while you are making decisions. You are never going to have 100% of the data that you need in order to make a decision, but you have to make a decision. So it's important to make those decisions sometime based on 70% of the information that you have. And sec, is that thinking big, not thinking incrementally, but thinking not 10%, but thinking 10X. And you may not land with 10X or 100X, but you're probably going to be much better than 10%, if you think 10X.

Ben Taylor (35:05):
Yeah. I think thinking big is much more exciting. You only get one life, so you better... why not swing for the fences instead of take the next step?

Debanjan Saha (35:13):
Take the next step incrementally.

Ben Taylor (35:15):
I really liked your mention of meeting with founders and VCs. Because I imagine for some of these individuals that you're meeting with at these cafes, I would imagine your perspective is constantly challenged because these people they're very accomplished, they're smart, they're intelligent and they have perspectives that probably shape you. What are some of the most valuable lessons you've learned interacting with these types of individuals? And maybe that's a difficult question.

Debanjan Saha (35:44):
Well, I think of this as brainstorming session or spotting sessions, right? And I set up these discussions and they have their own ideas. Sometime they're telling me about the investment they have made or seeking my ideas about the investment they're planning to make. And sometime they're challenging me about what I'm doing in this job or previous job, what the company is doing and if that is the right direction or not. And the way I think about it, that there is no right or wrong, right. These are all data points that you get through your interactions with various different people. And based on those data points, it's almost like a machine learning training exercise. You have all of these data points coming at you and you have to figure out what is the car, what the plane that you're drawing, which will give you to the next decision point that you are seeking. Right?

Debanjan Saha (36:51):
And ultimately, there are probably many rights and many wrongs. There is not single wrong or single right. These data points are very, very helpful in getting a multitude of opinions and ideas and then synthesizing what you need to decide. Right. That's the way I think about it. And in my job also, I mean, this is what I tell people I work with also, my job is to get... in a current job, for example, there are a lot of very, very smart people that I work with. And I typically don't try to decide myself. I get everybody involved. I facilitate the right discussion. And I think my job as an executive is to synthesize the best judgment out of the collective wisdom of the team. And if you expand that notion, trying to cultivate or trying to harvest the collective wisdom of the professional network, that I have to synthesize the best judgment that I can make.

Ben Taylor (37:54):
That makes sense. I love the technical founders. I'm curious if you see this for when you're doing diligence, do some technical founders fall prey to loving their technology too much? They care too much about the how and they don't really communicate the business value? Is that a theme you see with early founders?

Debanjan Saha (38:14):
I see a lot of that. In fact, a lot of the coaching and mentoring that I do is to focus them on the go to market part of it, the product market fit. And again, the foundational aspect of that is really focused on who the customer is. What is the use case that they're trying to solve? And is this going to fundamentally improve their quality of life or quality of business or whatever it is? Not by 10%, but 10 X. Right? There are two types of founders that I see or projects that I see, one where it is really an execution risk they're taking. There is already a product that exists in the market and they're trying to build a better product to replace the current product or current infrastructure or application, whatever they have.

Debanjan Saha (39:05):
And there, it has to be a significant improvement so that people will leave what they currently have to take a risk on something that's new, right? That's one type of problem that I see. The other is where they're trying to create new market. There is no existing demand, right? They think that they're building something which will create that existing demand, which is much harder. And you have to be a lot more selective. But if you are successful, the impact from that is actually much more than replacing your with current demand with a new product, which is easier to do, but probably is going to be... What should I say? It's not easier to do. There is more competition in that space. Impact is probably lower than creating a new business or creating new demand, but 90% of the cases I see where it is really replacing existing demand with a new product.

Ben Taylor (40:03):
And interacting with founders is fun. I've said something before that maybe sounds a little bit more critical where I said first time founders are a little silly because they believe they can change the world. And I had an investor tell me, they need to believe that because if you have this irrational hope that your startup could change the world, then that will help you weather the storm, the guaranteed storms that are coming. And so I thought that was good feedback. Is that also contagious interacting with these founders where they-

Debanjan Saha (40:34):
Yeah. It definitely is. And I do think it is important for founders to believe that. If they don't believe that, it's very difficult for them to change the world.

Ben Taylor (40:50):
The next question I have for you it's a fun one. It's a personal question. So your wife is also in the AI space as well, is that correct?

Debanjan Saha (40:58):
That is right. Yes.

Ben Taylor (41:00):
So what do you guys talk about over dinner? Do you chase AI questions down the rabbit hole? I'm curious what that dynamic looks like.

Debanjan Saha (41:12):
We try not to talk about. As a family, and not only my wife, my son is also very interested in AI. He's high schooler and he has his own GPU machines and he's training his models on that. My daughter is in vet school, by the way, just so that you know, and tend not to talk about AI at dinner table. Me and my son actually talk about that. But between me and my wife, we try to keep the discussion away from AI. But as a family, we are totally invested in AI. We do have some diversity because we don't work in the same company. That's a good investment, but we talk about other things.

Ben Taylor (41:50):
That's going to be great for your son to get different perspectives, where he's getting the best of both worlds. I'm curious what advice would you give to your son or to your younger self for people that are starting their careers early on? Because it's definite you wouldn't advise them to go to networking now. What advice would you give to the people just starting out?

Debanjan Saha (42:11):
Yeah. Well, I think what I have learned hindsight is 2020, of course, when you are, for example, getting into a college or getting into a graduate program, you're probably going to graduate four, five years from now, or if you're joining a job, you're probably going to be in that career for another 15, 20, 30 years, it is important to imagine where the world would be five years from now. And then pick an area where you think you are going to get a natural lift from overall growth in that area. That makes life a lot easier. Not that you should not try to do other things based on your passion, where you are trying to create a new discipline or going against the grain in order to make your point. Those things are fine also.

Debanjan Saha (43:01):
But for vast majority of the people, if you at least have some sense of where the world is moving, either yourself or talking to your mentors, your sponsors, people who probably know, have a better effort share into where the world is going, I think that makes life a lot easier. Not that that's my advice to everybody, making life easier, but I do think it helps. Second thing as I mentioned before, think big. Don't think that you're going to do small things, think big and think 10X, 100X, not 10%. And depending on where you are in your career.

Debanjan Saha (43:40):
I think other thing that I have seen work well for people, including myself, is taking risk. And the risk tolerance is different for different people. Risk tolerance is different also in different time in your life, in your career, but without risk, you don't make it big. Right? Of course, when you take risk, you have to make sure that you mitigate the risk of failure to some extent, but without risk, you don't make it big. And the final thought is learn and be curious. We are in a world where it changes very, very quickly. What you are doing today, you're probably not going to do five years from now or even three years from now. So it is important to learn every day on your job.

Ben Taylor (44:25):
I really like that in the statement of learn and be curious. Because I think if people want to march up the AI hill, if they're fighting to go up the hill, it's very different than them being pulled. And so if you're curious, and if you have a passion around it doesn't feel like work, but if you're trying to run up the hill for income, it sounds like a good job to have. That does not sound fun to me.

Debanjan Saha (44:48):
That's right.

Ben Taylor (44:49):
Well, I think we're wrapping up. We're coming to the end. Are there any questions we haven't talked about yet? Any themes you would like to hit on before we wrap up that you think would be helpful?

Debanjan Saha (45:01):
Well I think one thing I have been thinking about, I have been in the industry for a while. I have seen the internet revolution being through the data stack and ultimately hopefully AI/ML. So what's next? What comes after AI/ML 15 years from now? What is that we are going talk about?

Ben Taylor (45:24):
That's a very hard question. Some people think that AI is the end. That AI will reach a point of human intelligence where it can do any task that you decide you don't want to do. So if you don't want to mow your lawn, you don't have to. If you don't want to clean your house, you don't have to. If you don't want to read your emails, it'll give you the executive summary that an EA would've normally done for you. So it's hard to know what's next, but I guess there are verticals that we see. There's breakthroughs in healthcare, genetics. Hopefully, we'll be alive to see humans live on Mars.

Debanjan Saha (46:04):
I think that's probably going to happen.

Ben Taylor (46:05):
Yeah. Such a exciting time to be alive because I think there's so many innovations that are coming that we would've been convinced were science fiction, especially a decade ago, you would-

Debanjan Saha (46:16):
That's right.

Ben Taylor (46:19):
Yeah. Some people believe AI is moving... it'll move so much faster in the next decade than the decade before. If that's true, the innovations that are coming are hard to imagine, all of them.

Debanjan Saha (46:33):
I believe that. I believe that. And even in the last 15, 20 years, I think innovation has gotten faster and faster. And the cycle of innovation is getting shorter and shorter. And it's also creating value to not only people in the first world, but also everywhere in the world. People are seeing the value of that, right? You will find internet in every cafe around the world, which is not the case, even with telephone 15, 20 years back, right? If you go to any part of the world, everybody is carrying a cell phone and they have Facebook page or Instagram page of that. Right? So it's an interesting time to live in this world and I feel lucky that we are in the middle of it.

Ben Taylor (47:19):
I agree. I think the final question I'll ask, this podcast is called the More Intelligent Tomorrow, I think we've already hit on it a little bit, but what is the future in your imagination that you are the most excited about? What is an innovation that could exist, doesn't have to be AI related, 10, 20, 30 years from now?

Debanjan Saha (47:40):
Well, I was trying to ask you that question. I wish I knew. I do think there are a couple of interesting things that's happening. I do think a world which is greener is something that I'm really excited about. I think that's going to change a lot of things in this world, starting from infrastructure all the way to how people live. I do think the human being is probably going to get out of art and go to other places. It's difficult to probably see this right now, but it has to happen. Otherwise, the risk of living on one planet I think is too high. Yeah. But I don't know if that's going to happen in 30 years, but I think that's going to happen.

Ben Taylor (48:26):
Another thought that showed up with a previous guest is supply chain is depressing sometimes. So my nine year old wants a Nerf gun. Most likely that Nerf gun was built in China, came across to the cargo ship, was put on a truck. And so if you look at the carbon footprint of that Nerf gun, it's not ideal, but you can imagine a scenario where I would've 3D printed that Nerf gun and my son would've played with it. And then when it broke in four months, I would've melted it down and put it back into the process. So where we can have that experience, but it never left my house. So I think that would be a very exciting feature if something like that could come to pass.

Debanjan Saha (49:09):
I can see that happening by the way. I mean, that probably is going to happen. Just not a gun. You have a recipe, you want to cook your food, you just go there in front of your machine and then "print it."

Ben Taylor (49:18):
Yeah. Or you have the 3D printing, but the thing I saw that showed up recently is you can also get it's synthetic meat. So they're essentially 3D printing tiger steak or these different things where it's grown in a Petri dish or a lab and that... And so this feels very Star Trek... where maybe in the future, we won't actually eat something like a chicken leg. It's just 3D printed for you.

Debanjan Saha (49:43):
That's right. A lot of exciting stuff still to happen. I'm always confident that human ingenuity [inaudible 00:49:51] will make the day for us. It's difficult to imagine what's going to happen 15 years from now, but I'm pretty certain that it's going to be more exciting than what it is now.

Ben Taylor (50:00):
Yes. Which, just that thought alone is very exciting. Because things are interesting today. I'm curious, are you a sci-fi fan? Do you like science fiction?

Debanjan Saha (50:08):
I used to be, I was actually an avid static fan. I don't watch it that much anymore, but I do read about these days when I'm on my phone at the airport or things like that, I read up something quickly.

Ben Taylor (50:23):
It's interesting how many things from Star Trek we see today.

Debanjan Saha (50:28):
That's right.

Ben Taylor (50:28):
Things like the video, the live video chat on your phone, something that felt impossible is taken for granted. Well, Debanjan... Sorry, tongue time. Great thing about this not being live. Debanjan, thank you so much. Fascinating talking to you. Really appreciate your perspective and like you, I'm excited for the future to see where we go.

Speaker 2 (50:53):
Thank you for joining us on this More Intelligent Tomorrow journey. Discover more and join the conversation at moreintelligent.ai. The future is closer than we think.