SDS 169: Data Science: Technology and Philanthropy

SDS 169: Data Science: Technology and Philanthropy

Technology and PhilanthropyWelcome to episode #169 of the Super Data Science Podcast. Here we go!

It’s time to use your knowledge and skills as a data scientist for a greater mission! On today’s episode, Tarry Singh talks about how important philanthropy is in a company, how he changes the world for the better through technologies, how to make use of data sets effectively, and many more!

Subscribe on iTunesStitcher Radio or TuneIn

About Tarry Singh

Tarry Singh is the CEO, Founder and AI Neuroscience Researcher of an AI startup

Deepkapha focuses on the following three pillars: exclusively breakthrough AI research that intends to knit the world of neuroscience, models and frameworks around deep learning for the future, advises management and trains enterprises to holistically build AI departments with hands-on, market-relevant advanced AI projects and philanthropic initiative “givebackAI” to train the world that cannot afford expensive education and also works with organizations (Re:coded & Think.iT currently) that help disaffected youth in war-torn or ignored areas such as Syria, Tunisia, Iraq, Jordania etc.

Deepkapha has recently collaborated with IBM to work on building a holistic framework to unify various deep learning frameworks. This was announced by IBM in its global Think even in Las Vegas on 20th March 2018. (Announcement is here:

Tarry has over 17 years of experience working with data and has advised CxOs of global organizations to setup data-driven organizations from scratch. He speaks regularly at global AI leadership summits worldwide and conducts workshops on a regular basis with his TAs who are currently PhDs in various above-mentoined disciplines. He also participates in co-supervising PhD projects related to above areas with world’s leading universities such as Charité in Berlin.


Tarry Singh’s Deepkapha platform will launch soon! And he’s giving us a quite intensive peak on how his brainchild start-up company could change the future of industries and also, the entire world. So, how does Tarry plan to make this vision visible and powerful to reach everyone? He says that Deepkapha is rooted onto three essential components: DeepRRP (the research arm), Enterprise AI Engineering Advisory (the business arm), and Philanthropy.

Before he incubated the idea for the start-up, he recognized two evident problems while he was traveling the world. He noticed that there is a shortage of engineers and there is available talent that enterprises are not taking advantage of.

Tarry emphasizes most in philanthropy since this is also one of his greatest passions as a humanitarian. Make your purpose bigger. Tarry shares with us inspiring stories about how he gives back to the data science community by reaching out to the less privileged around the world, especially in the third world countries. Kirill also shares that emphasizing on the philanthropy arm is also essential in his efforts on bringing forward data science education. Bringing tools and techniques in AI, machine learning, and deep learning could help third world countries’ economic status drastically.

On the next part, Tarry discusses the difference between Hinton’s capsule theory from the traditional deep learning. He also shares at least three research papers he’s working on right now. He also advises making use of data sets to make better algorithms for better performance.

There are five main steps or ‘plateaus’ as Tarry calls it, for you guys to take when you want to start in deep learning and AI. No need to fuss when you don’t have the data science background. Tarry will guide you in a step-by-step way to not get overwhelmed and be an expert in the field. Tarry says that you don’t have to be perfect in order to move on to the next plateau. Just remember to go back to the 1st steps and the other steps before to be updated on new knowledge. Along the way, you could choose what appeals to you!

There’s a lot more to learn from Tarry especially in his philanthropic efforts through machine learning, deep learning, and AI. Start tuning in!

In this episode you will learn:

  • Tarry Singh looks back on when he stopped doing what consulting firms are doing and finally decided to delve deeper into deep learning through (05:20)
  • The two problems of industries that Tarry has observed while traveling around the world: there is a shortage of engineers & there is available talent but not getting connected to the industry. (07:19)
  • Three Components to keep in mind when setting-up for companies: Research arm, Enterprise advisor, and Philanthropy. (10:00)
  • Kirill also weighs in how important the philanthropy component is as seen in his online courses. (14:05)
  • Teaching about tools and techniques on technologies (e.g. deep learning and AI) to people in third world countries could change their economic reality dramatically. (16:05)
  • Take some bold steps, if necessary, like talking to governments to bring forth learning and expanding the knowledge about the technologies. (22:34)
  • What is Capsule Theory and how is it different from traditional deep learning? (24:40)
  • 5 Main Steps/Plateaus to climb if you want to pursue deep learning. (34:32)
    • 1st Plateau: Fundamentals (37:05)
    • 2nd Plateau: Visualization (38:00)
    • 3rd Plateau: Machine Learning (38:36)
    • 4th Plateau: Deep Learning (39:48)
    • 5th Plateau: Applied AI (40:30)
  • Make use of existing data sets to make better algorithms for the better world. (46:00)
  • How would someone know that it’s time to forward from the current ‘plateau’ you’re standing on? (52:10)

Items mentioned in this podcast:

Follow Tarry

Episode Transcript


Full Podcast Transcript

Expand to view full transcript

Kirill Eremenko: This is episode number 169 with Data Science Thought-Leader, Tarry Singh.

Welcome to the Super Data Science podcast. My name is Kirill Eremenko, Data Science coach and lifestyle entrepreneur. And each week we bring you inspiring people and ideas to help you build your successful career in Data Science.

Thanks for being here today, and now let's make the complex simple.

Welcome back to the Super Data Science Podcast, ladies and gentlemen. Today I've got a very special, very exciting guest, Tarry Singh, who is a founder, a CEO, and AI researcher, a Data Science Executive, a philanthropist, a speaker, and just a very, very nice person who gives back so much, so, so much back to the data science community. Who educates, who helps people, and I was very honored, very grateful to have Tarry on the show today.

We had a lovely conversation, and we just let it go where it went. We had no idea what was gonna come out of it. We talked about things like philanthropy, we talked about data science education, helping people out, where the World is going in terms of getting third World countries on the Tech radar, and helping people in less privileged societies get up to speed with data science. And what contributions we as individuals can make towards those causes.

We also talked about Jeffrey Dentons recent capsule network and capsule theory. So if you're interested in that, then this podcast is for you. We also talked about the research that Tarry himself is doing through the research arm of his business. We talked about advising executives and enterprises on data science, and how all of those components come together. We also talked about Tarry's recent major infographic hit that is very popular on LinkedIn. He'll actually walk us through it, it's called Climbing the Hill of Deep Learning. But it's actually not just about deep learning. It's about the whole process of building your career in data science and exploring different opportunities, and those five different plateaus at which you can build your career. So you'll get Tarry's advice straight from himself, from his experience, and from his expertise in working with thousands of data science students and data science professionals in person.

So there we go, that's what today's podcast is all about. A very lovely conversation, exciting journey. Can't wait for you to join us, let's dive straight into it.

Without further ado, I bring to you Tarry Singh, a data science thought-leader.
Welcome ladies and gentlemen to the Super Data Science Podcast. Today I've got a very exciting guest with me on the show, Tarry Singh. Welcome Tarry, how are you today?

Tarry Singh: Thank you Kirill, thank you for having me. It's great weather here in Amsterdam. And I'm super excited to be part of your podcast show. Thank you once again for the work that you have been doing tirelessly in the last couple of years for data scientists. I think we all know who you are, and I'm very thankful and grateful to be part of this podcast.

Kirill Eremenko:: Thank you, and same here. Just recently it's been interesting how your name as been popping up, Tarry, Tarry, Tarry. And I am also a very big fan of all the contributions you've given to the world of data science. All the wonderful materials you've created, all the advice, and insights that you've shared back to the community. So, very excited about this chat. As we discussed at the start, we don't have any predefined agenda that we wanna talk about like plan how we're gonna go through this. Just let it flow and see where it takes us, right?

Tarry Singh: Absolutely. I mean we are all in the same field. Data science is expanding actually in all directions. And I think in the similar way the conversation will also lead to our intuitions, which I hope the audience will be able to enjoy as well. So let's keep it free-flow, yes.

Kirill Eremenko:: Sounds fantastic. All right, let's maybe start with your company. So you're the CEO and AI researcher at, and I'm happy I pronounced that correctly from the first time, as you said.

Tarry Singh: Absolutely, yeah. You're one of the very few people who has no problem at all in pronouncing.

Kirill Eremenko:: Yes.

Tarry Singh: Thank you.

Kirill Eremenko:: Yes. All right. Tell us a little bit about deepkapha, what is the company all about?

Tarry Singh: Cool. I've been in this industry for like 25 odd years. About a year and a half back, almost nearly two years back, I decided that I did not want to be part of a consulting world in which I sort of pretended that I knew what I told my customers. I just needed to take a break to get deeper into deep learning. I'm being very honest here.

The reason why, is not taking a jab at the profession that I've been previously in. It's just a field that is expanding dramatically, in giving back to sort of what you guys have been doing. Your podcast series, and your educational series may have been very educational, but they also opened up a huge new world of data science. Now, when I look back about almost two years ago, I said, "Okay, let's stop and let's go deeper into it."

I had already established two companies in the past. My first start up was a management consulting start-up in which we wanted to sort of break the bank and do some amazing things. My second start-up was an NLP Social Analytics back in 2012. So I was already playing around with this, but didn't realize that I needed to explore myself and explain to the World myself.

So I traveled around the World last year, in the beginning of the last year. I met some global world leaders who've been running some billion dollar companies, tech companies, and also met and also interacted with people in Montreal, in Toronto. And also, [crosstalk 00:07:07]

Kirill Eremenko:: We all know who you're talking about in Montreal and Toronto at University of Montreal, Geoffrey Hinton and company. That whole ...

Tarry Singh: Yeah. So just kind of disclaimer ... So these are the kind of interesting conversations we've been having. For example, Geof Hintons paper which they released in capsule. SO many conversations, some conversations are very intense, internal. But also industry leaders, guys who've been running big companies, internet companies in China, also here in Europe. What I realized was that, I think there was two things I realized. One was that there is a huge shortage of engineers, and I foresaw a huge shortage of engineers. We were obviously aware of this trend that Google, Facebook, and all these other companies are constantly getting the best talent from Europe and all over the World. All the Masters and PhD students in different areas in healthcare, or bioinformatics, they're all moving into these big companies.

It's leading to a huge problem in the industry. I knew this because I come from the industry, I've been there for a long time. And the second thing which I've realized, when I was traveling and making these travels around the World, I was giving speeches and conferences and key-noting ... that there is talent available, but it's not being connected to the industry. So what's going wrong? I decided that I would create a silk route ... I'm calling it an AI silk route, that's also part of my pitch to the investors ... That I will work with these people, I will start giving workshops, and bring these people to the industry. Because the industry leaders constantly ask me, "Hey Tarry, we wanna set up and AI lab, we wanna set up, do this, do that. How can I do this?"

It's very hard to get the right talent to get started, So by the end of last year, I was already getting some offers from a chairman of a large 25+ billion dollar company. He reached out, and I started this project. Then I realized, oh God, I don't have an entity. I was incubating this idea so we incorporated the firm, calling it deepkapha. Deep for Deep Learning. When I say deep learning it's .... Deep learning is deep reading and deep understanding. I didn't go into this technology concept which is so popular right now. And Kapha is more about harmony. How do you bring these two together in a harmonious way so the World can learn together?

Long story short, since January when we incorporated, until now, I decided to ... I said, "Okay so I am starting a company, why not do it the way I always wanted to do it since I was a kid." I wanted to learn and play. So I said, "Okay, then I'll set up a research arm." I wanted to continue to stay in touch with the reality, which is the business World out there. Because these are the guys who need AI people right now, right?

Kirill Eremenko:: Yup.

Tarry Singh: I mean, we cannot just keep promising our people, our young engineers, that there's a place for you in Google or Facebook. These companies cannot continue to keep taking hundreds of thousands of people. They also have sort of a stop sign somewhere, saying, "Okay, no more."

So I said, "Okay..." Since I have worked with enterprises and advised chief executives of large companies for quite a while, I said, "Okay, so this is a nice conversation I can have with them." So I decided to set up an enterprise advisory for AI as one business unit. The other is research, and the third, which is far more ... sort of appeals to me as a human, is to really do it selflessly. How can I do this from philanthropy perspective? Because there are many people, smart people who don't have money. These are very bright people ... kids even, very young kids, 12, 13 year olds who are planning a future, who read a lot of books but somehow don't have funds.

I also reached out and I was also approached by companies like Think.iT in Tunisia, amazing group of people there. A company called Recoded, which is a humanitarian firm working in Iraq, in Syria, in Turkey. So I just started traveling, going to these places together with them ... Obviously these were our partnerships, and also we had full advice and guidance from United Nations, it's still going on.

So this way, it was giving me a lot of satisfaction to do my job. Because normally happiness is a difficult thing when you start on your mission, and you have to deal with the hardcore world which is either enterprise. So this way it gives me energy, but also keeps helping me bring more and more people into this world. Which is great, because that's the mission we have, right? You also have the same. How do we bring people, and more people, so we create these ... It's almost like saying ... you go in front of this big castle, and you say, "Okay, so you know, hey big castle you advised that this AI is going to be shaping the new industry, and here, I have a few millIon people standing with me. And we want to enter, and we want to explore, and we want to make It much bigger."

So that's the way it feels. I'm not the only one fortunately. You guys are also in this game. It only helps us expand this ecosystem more and more. So Enterprise advisory to bring these guys some advice and get them to hire smart people. Research has been writing breakthrough research to write new activation functions, to improve capsule theory into much more detail, I can explain maybe later. So we are publishing papers that are going to improve the deep learning ecosystem literally, from algorithm perspective. And third is philanthropy, which is ... My heart totally warms up every time I have this mission, I have to go somewhere. So I said, "Okay, let's do it." You know?

Kirill Eremenko: Mm-hmm (affirmative)-

Tarry Singh: So let's say in a nutshell what deepkapha intends to do.

Kirill Eremenko: Fantastic. That's so interesting. I'm listening to your story, and you broke it down into these three components, and I'm actually seeing myself so much in that. So you mentioned Enterprise Advisor to help companies get these talented people on board. Research arm, to improve the ecosystem, and the philanthropy, because that's the ultimate mission, that's what gives you fulfillment.

For me, so similar. I'm actually so surprised. We started with this philanthropy component. I'm not going to go out there and say I'm doing this all just for philanthropy reasons. Of course it's a business. It has to grow, it has people that work in it. But at the same time, if you look at our courses, people studying, learning, can get these courses at such low prices, that's why we have hundreds of thousands ... We just crossed half a million students. And that stands to show that people really do want to grow and expand in this area.

I would say that component was our starting one. And then, funny enough, the research arm and Enterprise Advisor, we just launched two new businesses. One is a research business called, where we do research on new algorithms in artificial intelligence to help ... also expand the space and empower businesses-

Tarry Singh: Amazing.

Kirill Eremenko: ... to do more. And the other one, Data Driven Executives, is to help executives understand better how to become data driven and build these different companies. So, also Enterprise Advisor, it's like, your three points, I just check them off as well, so interesting.

Tarry Singh: Yeah. It's beautiful. The more enterprises and firms like yourselves, the more of these are in the industry, the better. I think it's really great because we need to go to Africa, where you are. I've been getting a lot of requests already from Uganda. I've done a project in Uganda a few years ago, that was 10 years ago. So I think Africa is a huge continent where we can have hundreds of thousands of people trained, maybe millions. I think we need more guys and more outfits like yourselves, so we can create this ecosystem and make it much bigger. So amazing, I'm really happy to hear that you're doing this as well. Amazing.

Kirill Eremenko: Thank you, thank you. And just on that point ... cuz listeners might be a little bit confused, I am in Africa. I'm just here on an island in Comoros. To your point, it's a very far away place from everything. And its kind of in the middle of the ocean, and there's a lot of poverty, it's a very poor place. It's one of those place that ... It only exists because of a certain industry, in this case it's Chanel No. 5 that export this plant called the Ylang Ylang.

But amidst all this poverty ... Like today morning I went jogging on the beach, and I saw one of the local kids, or maybe my age so I won't say kids, maybe young adult, and he was also jogging. And he had a phone, like an iPhone, and he was listening to music. So even though there is so much poverty, they have access to internet. The World is so different to what it was 20 years ago, even 10 years ago. They have this access. And by empowering people with online education, sharing online knowledge and these things, you can really change their lives drastically. It just gives them a little bit of inspiration and they will embrace it, and they will soak up all this knowledge and change their lives.

Tarry Singh: Yeah, absolutely. I can just add one thing to it. I was interviewed, I think two years ago, or was it three years ago, by a journalist who used to work for Al Jazeera back then. He was looking for a story ... We had this conversation, it was published by a start-up, a french start-up, I forgot the name. It was like a news aggregator kind of a start-up in which they aggregate news and make some interesting stories out of it. So he asked me, "So what do you mean about technology to get pervasive?" And I gave a ... from policy and from migration perspective, which I still very strongly believe in. I think the problem that we are having today in Europe is it's .... essentially crisis for European Union. You have boats floating all over the place and Italy doesn't want it. Spain, for example, yesterday you had this problem with hundreds of young people.

I look at the boat, and I'm seeing those young men struggling. These are like 15 to 25, young African men and kids. No one in this World Kirill, wants to sit in a boat and go to some country which is strange, no matter how wealthy it looks, and eventually end up on street. Or never be able to get that job which you actually really deserve.

I spent four years in Uganda doing a project through Dutch Ministry. I don't talk much about it, but I'm very proud of that project which I did to bring awareness, but also spread technology. I believe that if we start bringing technology where people can start building businesses and start doing things, they would be so great. They would set up their own economical ... Their economical reality is gonna change dramatically. They are not gonna look a those boats and make those horrible and dangerous passes to come to Europe.

I think it's a win-win situation if you bring deep learning and artificial intelligence in its own beautiful way to other parts of Africa. For instance, Kenya, Uganda and even Rwanda, is really improving. As you know ... you may have heard yesterday, day before yesterday, the reason why they announced that they wanna sponsor arsenal football club with donation is because, they say, "We wanna get rid of the money that we get from all these other richer countries." Because it's a stigma. All these countries, even from Netherlands, it's like 45 million or something, or maybe more, that goes into Rwanda. So these guys are saying, "We don't want your money. I want to build my own nation."

From a policy perspective, it's great to give people tools and techniques. And I think Africa is going to be the huge, huge continent the World should be looking at, really. From expanding this knowledge.

Kirill Eremenko: Yeah. Exactly. Have you heard of Peter Diamandis' X Prize for education? The one for Africa?
Tarry Singh: No I have not. We are working with Think.iT, and I know Obama, Barrack Obama, the U.S. President, he's also launching a fund for Africa. And we are in conversations with the founders of ... the CEO of Think.iT. They're amazing people. So there are some conversations going on to start that. But Peter Diamandis, I know he's invested in a company of a gentleman I know in Boston. But I haven't heard of this initiative, no.

Kirill Eremenko: This one is very similar to what you're describing. There's a prize, I think it's maybe a couple hundred thousand dollars, maybe up to a million, I'm not sure the exact amount. But it's about ... Or maybe it's actually already finished, cuz last time I checked on this was about a year ago. But anyway, it's about creating an application for iPads in such a way that anybody can pick up this iPad and learn basic schooling things like Mathematics or English, or Geometry and things like that, without any guidance. And so basically, the plan is, as soon as that app is developed, and tested, and it's verified, what they're planning to do is to drop several thousand, or hundreds of thousands of these iPads throughout Africa. And just leave them in different places so any child can pick it up. And by clicking, without any guidance, without understanding the language, can actually learn new stuff. How cool is that?

Tarry Singh: Yeah. I remember in 2006, when I started this project in Uganda. There was also an initiative called OLPC, or One Laptop Per Child. I'm sure you've heard of it as well.

Kirill Eremenko: No actually, I haven't heard of it.

Tarry Singh: So it was ... I know I carried this as well, in fact I used to bring it also to Europe for conferences here, back then. So I think in a way it's similar, it's interesting actually. The thing is these things need to start rolling.

So that is one. What I think [inaudible 00:23:07] from experience I can give a word of caution is, it is people like yourselves, and myself and others who need to go there and bring this education in the classical way. We should not forget that in European or other modern economies, young people can sit behind a computer and learn. While in Africa ... but also other Asian ... Really you are from Australia right? So you know in Asian cultures, people like to sit together and understand it from a community perspective, and also physical and classical perspective. Someone standing and teaching me.

I think culturally anyways, but I think we have to take some bold steps to set things up and maybe work with governments if necessary. And that's what we are exploring now in Africa, from a country governmental perspective, to bring it in a more holistic way. And expanding it in a way that people not only listen about it, but they think it, and they can then expand it. And I think that is needed.

I realize in those four years I spent at Uganda, bringing in technology, starting prizes has a spiking effect. Not like a neural spiking. But it's interesting when it's there. But the minute it's gone, people go back doing the same things which they were doing. So that's the danger we should be careful about.

Kirill Eremenko: Gotcha. Thank you for that discussion. I'm sure there's lots that we can all do in that space. Now, let's move back a little bit and talk about deep learning and some of the recent developments in that space. Specifically I think a good place to start would be capsule networks. So, I don't know much about Capsule Theory, which Geoffrey Hinton released recently. I know there's this one medium blog post which is pretty popular on that space. Could you give us an overview? What is Capsule Theory, and how is it different to traditional deep learning?

Tarry Singh: Cool. So capsule definitely is a hub for many researchers. Geoffrey Hinton in fact wrote a paper back in 1981 in which ... In fact a few wordings which we see from capsule's paper, there are some quite similarities with what Sarah Ward, one of the authors, has written about it.

First of all, it's not really that new, the whole concept of poles, rotation, and basically trying to understand the sparse or limited data about us let's say in a certain manifold space. Meaning if I look at Kirill from side, I just see part of his nose, or eye, or things like that, then I understand that it's Kirill, I don't need the MSCoe code ... huge dataset to [inaudible 00:26:13] to figure out it's Kirill.
The rotation of your head, even the back of your head, I can very quickly say, "I think it's Kirill."

Kirill Eremenko: Yeah. Whereas A-

Tarry Singh: I think this is what we are trying-

Kirill Eremenko: AI can't do that at this stage, right? Deep learning-

Tarry Singh: No.

Kirill Eremenko: ... can't look at the back of somebody's head and say it's Kirill.

Tarry Singh: Exactly. So it's almost like you have this, maybe year and a half year old little kid, that kind of sort of cognitive capability we have helped AI achieve. But it's not moved beyond that one and a half year-old kids cognitive capability, who is maybe a little bit drunk and not being able to see things at once. Things look different, or tilted. So for example, if there is a tilt, if there is a pose and change of rotation, texture, warmth, and different things that are attributes to who we are, and we put the three-dimension into it ... from the three dimensional perspective, and then also start adding different attributes to the same in which I see, for example, Kirill, then I should still be able to make ... So for example, you're in Africa right now, there might be a gazelle flying on top of your head. As a human, even as someone who has not seen that data, for example. This is a new data that has actually been created in my brain, I'm able to make full sense out of it.

So capsule theory, basically what it tries to do is, it's trying to mimic more in a away in which how neuroscientists have tried to understand how the neurons are firing inside of brains, how they are grouping together. So this whole idea of routing by agreement is more about ... sort of, that's the algorithm, part of the algorithm which tells the network that, "Okay, so we agree as a group of neurons that this is what it is, irrespective of everything else that I see around Kirill's environment which is strange, is this huge, weird marshland. It's not a hack, it's real because I know he's there, and then this honey Gazelle which is two meters in the air." It's something which I can correlate to a certain extent and say, "Well it looks strange, but we agree." And then the neurons basically ... you take those neurons and you pass it in the apex, and eventually try to make more sense out of it.

Having said that, I think this is the beginning of what AI should become as we move forward improving this network. It's very relatively new. It takes about two years, if you look at the experience with the convolutional network also, before the accuracies are improving in other ... Let's say improvements are being brought to the network, which we are working on as well.

I'm very happy to share with you in brief, because we haven't published those papers yet. But we are trying to bring more automated and more intelligent algorithms into the network, into the neural network ecosystem. But all in all, basically it means is trying to understand from three dimensional, trying to ... hopefully with as limited data as possible to make approximations which essentially, as you've seen, you've heard of the pixel attack and all those things, the convolutional neural network ... Hopefully we can move into more intelligent and more human-like and neural network.

Kirill Eremenko: Gotcha, gotcha. And please do share, I'd love to hear some of the ... whatever you can, some of those research papers that you're working on. What's the most exciting thing that's happening right now for you

Tarry Singh: So right now we're writing three or four papers. One of the papers is actually being released, two papers. So my goal is through a research arm, I have a head of research, she is a neuroscientist, she's completing a PhD here in Berlin, we just published a paper in ICSE, it's called ICSE, 40th software engineering conference in Gothenburg in Sweden. So [crosstalk 00:30:26]-

Kirill Eremenko: Gothenburg sounds like back then, right? Didn't know that was a thing. All right, sorry, yeah. [crosstalk 00:30:35]

Tarry Singh: So we presented our paper, Neural deals. We're calling it the Neural deal, meaning trying to use as much of neural science data collection which passes through from our retina to our neocortex. What we are trying to explain is that there is a lot of data and a lot of data manipulation that happens between these two junctions, meaning your retina and the back of your head. And then how can we use this data to basically start maybe creating new different algorithms, for example, back propagation probably is still rather immature. However great it is in making approximations today, it's still not the realistic way of how we, let's say, deduce information about the world.

What we are writing is ... We're improving a squashing function. Which is the activation function which capsule networks has. We are calling it an in-squash. So we are trying to, introduce a second order norm to it. We are still right now testing vigorously on our servers. The second which we are adding, which may not necessarily have anything to do with capsule, but obviously we want to include it into the capsule framework, is trying to bring a deep switch. We are calling it a deep switch internally right now. What it means is that we should be able to switch across various optimizers that are there while you are running your network. So you don't have to babysit 10 different networks with 10 different models. And you can just have it switched based on certain parameters and certain sets of conditions.

So that is one, and then we're trying to combine also other simple [inaudible 00:32:29]. It's very interesting, my researchers are setting it up right now. For example, even learning with hyper-parameters, which essentially we either just run our network or train it for weeks. But right now we are saying, "Hey, hang on. Let's kind of [inaudible 00:32:44], this whole learning rate." Be more adaptive. Make it more auto learn while it's running in the network, and make it more interesting.

And then there is obviously in capsules, we have already done some research applying manifold learning and unsupervised learning. And right now we are currently experimenting heavily on PGM, so probabilistic graph models. We are basically trying to force this whole unsupervised learning, as I just explained, Kirill in this strange, funny grassland, a gazelle flying on his head, over his head. In fact, I had a picture also on a research paper in neuroscience, Neurodeal, in which there is this car flying in a jungle. Very weird. So it's there in that illustration, and that paper should be going into archive very soon.

So those are the kinds of papers we are writing. And we keep talking to each other, because as you know, writing research sounds interesting from far, but there's a lot of research that fails as well. We have to accept that and move on and keep trying new models.

So there are four or five papers we are writing. We've already ... For example, one of my researchers has written a paper on an activation function which improves on ReLU and factors better than ReLU. So that has already been published, it's on archive. [crosstalk 00:34:13]

The research is really interesting. It's like kids coming together, and we start Playing Lego with each other.

Kirill Eremenko: Yeah. Well congratulations. All of those sound like very interesting, pushing the envelope type of undertakings. So excited to see what comes out of that.

I wanted to move a bit to the side here, and talk a little bit about ... More for our listeners who are just getting into this space of deep learning. So you have this wonderful, fantastic infographic which you shared. At least I've seen it on LinkedIn, probably other places. Gotten tons of comments, tons of likes, and I'm sure many people have been impacted by it. It's, How should I start in Deep Learning and Artificial Intelligence? Got five main steps, I'm looking at it right now, and we'll share it in the show notes. If you don't mind, could you walk us through these five main steps, and maybe give us your comments so that somebody who is a bit lost in the world of deep learning, but wants to get into it will have a very clear pathway?

Tarry Singh: Yeah. So basically, I called it hill climbing, and this was part of ... This was a result of the workshops and the training that I've been giving to enterprises and groups. Hundreds ... I think I've trained already eight and a half to, I think it's probably already nine thousand people.[crosstalk 00:35:32] These are all classical. It's not online, I go to places.

Kirill Eremenko: That's insane. Where'd you find the time?

Tarry Singh: Yeah that's a huge number actually. If I look back after, it's almost a year now. In fact, it is a year. In June I really started doing this, last year. So I've already touched almost close to nine thousand people. These are in-house. People are asking how do I do this.

So I started sketching it, and I sketched it for over a few months, because it was also my own journey. I said, "How can you just throw information in peoples face and expect them to learn?" It's very hard for a lot of people. If you have ... The couple of basic things is that if you have intuitions in physics and mathematics ... I studied physics first in University and then Nodical Astronomy. So basically, I already was very curious about this World as a physicist, and as an astronomer, thinking the Universe, the World. So basically very observant, and at least very curious ... Observant, I don't know if I was observant enough as a young guy. You have other things to do when you're having fun, but still observing.

But not everybody is coming from that background. People may have business commerce background. People may have some other intuitions which do not help them see the light. So then I started sketching it, and the first which are called, I called it plateaus ... When you should climb a hill, you have plateaus, almost like climbing Mt. Everest.

The first plateau is the fundamentals. I started revising, collecting information and data, and also easy to understand stuff. For example, there is a beautiful book written by a gentleman, and I just forget his name, but he has written two beautiful books on physics and mathematics, and he calls it No Bullshit Linear Algebra, or something like that. Also on physics. So I started giving people those kinds of books that help people seamlessly climb into that plateau, without being intimidated. Because back in high school you have ... It's a priority, as in you have information and you're just thinking, "Oh my God, so I don't think I can ever do it."

I sucked at [inaudible 00:37:51] in high school. So that is a difficult step. Then I tell people, "Okay, don't worry too much about it." You're in plateau one. Plateau two is trying to understand visualization skills. You are not maybe an analytics person, you don't feel like it, you don't think you can write and algorithm and share it. Don't worry about it right now, let's start visualizing, you're a visual person. So the plateau two I started calling visualization, and then I give them introduction into all this visualization libraries. And slowly, in very seamless and easy to understand way, we start writing code together.

When people get comfortable, I said, "Go back to plateau one and try to see what you understood there. And make some changes, come back to plateau two, which is your visualization thing, and then let's move on to the third one." And machine learning becomes the third plateau. And there you have a ton of those series yourself which you guys created. So I point those, I point to several other areas, I said, "Look there, look there, look there." And make combination which suits the best, and try to keep your learning curve measured. Be honest with yourself. If you don't understand, go back and read it. If you don't understand go back to the plateau two, plateau one, come back again. So it's almost like going back and forth.

Once you master parts of machine learning, you don't have to do everything. So people start thinking, "I have to boil this whole ocean." So then I said, "Okay, just do parts of it." Maybe if you're moving into unsupervised, do support vector machine understanding, how Apne created. Get the historical perspective. Read why people made those things, why people wrote those things. That will help you remember these things longer than if you just remembered as a formula or some kind of algorithm.

So when people go back and say, "Okay yeah, this Russian guy, he created this and it happened this and he did that." And then they remember longer, then their intuitions start developing. When these things start happening, then I say, "Then you are actually ready for deep learning." Although I keep saying, "If you can already jump from plateau two to plateau four, which is deep learning, what is it then?" Then start showing them, explaining in a sort of easy to learn sort of way ... Going back to all intuitions, historical perspective. What was Boltzmann machines, and who was Ludwig Boltzmann, and what was his intuitions? What is the role of statistical mechanics? How does this apply to your activation functions that you're creating? And all these things.

Those perspectives start making ... It's almost like a story telling if you will. I think the fifth plateau is applied AI, which you need to eventually apply. Because people say, "Okay now I have every theory, I ran every darn dataset on Kaggle, and everybody's done the same. So I'm still ... It feels as if I'm part of the network in which everybody is saying the same thing. So what? What is my differentiation? What do I do?" And I said, "Okay." So that's the step in which you start looking at datasets. So go talk to your community. And then people say, "Well, it's easier to say." I said, "Hang on. It's not easy to say. You're right, it's not easy. So when you're going there, when you're meeting ... when you go into hospital, you have someone who is in the hospital network." Believe me, in hospitals, even in India and Bhutan, we're also ... I'm helping a researcher doing a project there. So there are people who are collecting data, all you have to do is just start going. People have data.

When you start going, then you start learning this whole art of data collection, pre-processing, creating balanced datasets. When you are starting to do that then you'll really feel like you're building something. You're almost like this guy who's building this brick house and used to go brick by brick, and I said, "You know this is the journey you have to go through and then you can reach that summit with applied AI." It could be anything, it could applying policy changes, it could be trying to change the way the world ... Income inequality, you can get statistical datasets from your country, and try to start making sense out of it.

I think there's a whole lot of things, and then you can start applying coronal network into some time series or something else. These things start helping. There are a lot of people who came back and are doing some really amazing stuff actually. So in a way, those five plateaus really helps you to really become a master in an area that differentiates you from your other peers. And this differentiation is eventually the trigger, or a catalyst, for us, for you and I to seize satisfactorily and say, "Well this is the network effect which we want to achieve when we mean that this ecosystem has to expand."

Because if we don't do this, I think the risk is that we will continue to train people in theory, and in toy datasets, and these toys are not going to make them real men. They are going to remain boys and girls. We have to make them men and women. Deep learning.

Kirill Eremenko: Yeah. Yeah, true.

Tarry Singh: That's my story, that's my opinion I guess. I have a bit of experience.

Kirill Eremenko: Yeah. Very clear. And I definitely agree with that. When I was creating the course on our programming, I remember I looked around the place and did some research of the existing courses. And one of the things that I noticed is that every single course out there uses the virginica setosa dataset. I was like, "Oh my God." It's so repetitive, right? Those flowers and the whole fisher iris dataset. And it's like, "Come on guys, we can do better than that." And I made it one of the core values of the course creation process, that I look for datasets that are current, relevant, interesting, from industries, real business challenges and so on. So that people learn through ... they can see that it's not just theoretic application to a dataset that was discovered 100 years ago. But it's actually something that is happening now. Something like, I don't know, some machines in a mining plant and you can predict their maintenance requirements. Or there's a consulting firm that is trying to help a bank differentiate or do something with its customers and segment them better.

You're right. By putting it into perspective like that, it helps people see that this is not just a theoretic exercise, I can actually make an impact. I can actually help businesses, people, charities, friends, organizations, myself, analyze and understand better. It inspires people to actually look at stuff. You can get your own Fitbit, or iPhone and measure how many steps you took and analyze that. That's already something cool.

Tarry Singh: Totally agree with you. I mean, make it real, make it practical, and make it stick in your head. Because it's not going to stick. Setosas and all these leaves or the MNIST and all these guys are not going to stick in your head because ... It's a great way to benchmark, so MNIST is a great way to benchmark if you've written a beautiful algorithm. But don't start using it as something to prove if you have to do 3D lung cancer, you need something different if you have that. I think we need also more advanced datasets that are normalized. That are presented to us in a way, where you have healthcare data, agricultural data, manufacturing data. There should be some interesting data sets coming which will help.

But I think that's a next way which we should be seeing in the next five years. You will have datasets for specific, all verticals that will help us get even better with our algorithms. So I totally agree with you. Yes.

Kirill Eremenko: Yep. Tarry, let's start a new business. Let's start a repository of all datasets.

Tarry Singh: I can tell you Kirill, seriously this is no joke. In fact this is one of the things also ... we are working on a patent as well. My mentor actually advised me that you need to go and file a patent. And it's all about datasets. Today we are looking at datasets and people are not making sense out of it Kirill. I didn't either, I was also like, "Oh yeah, yeah." Because you're focused on a mission, you're not looking around the world.

So, in their my mentor is amazing, he's almost like a second dad to me. He said, "Okay, hey listen, let's take a break. Let's go to a sauna, and you're not gonna talk about anything. I don't want you to start visiting up this big sequoia forest." I said, "Okay. So why..." I said, "No, no, no, I want to go there." And I had a keynote there in San Francisco with a bunch of people, Google, LinkedIn and all these guys.

So he says, "Let's go away." And we went. We spent the whole day doing nothing, and this is when the idea started coming to us. He says, "you know, you have all these datasets. For example, Google is releasing all this audio and video and all that stuff." And I said, "This is the new economy. The new economy is going to be based on the manipulations and even extrapolations and interpolations of these datasets. Because essentially this is what your brain does, right?" So I said, "Yeah."

Because I translate information in front of me which is visual, which is text, which is audio. And it constantly is transposing and interpolating, and that gives me intuitions. He says, "This is what the new economy has got to be. It's not just going to be in its own silo." The danger is that you will have companies like Google and Facebook< they will focus on their own silos, because that's where the business is. And there has to be someone who comes and starts looking from a horizontal perspective, and how do you create a cognitive layer from this ... This master algorithm thing, right, which Pedro Domingos wrote. What he meant was that how do we bring these five tribes together? But this whole idea of creating a master or supervisory algorithm, would be to essentially take advantage of mature datasets, which start teaching industries and verticals about their systems. And obviously you need an algorithm to run this, because the algorithm is the engine. But I think datasets is ... More people should be thinking about it. When I feel that I'm alone, I either apply Peter Thiel's formula that ... If there are just a few people who believe in it, and everybody else disagrees, you have a great idea. But my intuition says that I have a great idea, because I'm working on a patent, which I'm going crazy thinking about it. So yeah, why not? I mean let's have a chat. I believe we're also definitely going to be meeting in San Diego. That's something I spoke to your colleague ... Kirill Eremenko: Yeah, yeah. So for our listeners, I'm very excited to announce that Tarry is going to be joining us for Data Science Go 2018 in October, this year in San Diego. Super pumped about it, can't wait to meet you in person. How are you feeling about coming there and giving us a little bit of ... sharing some of your insights with our audience? Tarry Singh: Amazing, I'm so excited. Once again very thankful and grateful for everything that happens. So excited to meet you Kirill in person. I'm sure we'll exchange great ideas. I think it will be a great, great show. I spoke to Boe, Boe is a very good friend of mine. He is a kind soul, and I know it's such a successful thing. I'm very happy to help you expand. Because- Kirill Eremenko: Thank you. Tarry Singh: ... it's our common goal. Kirill Eremenko: Thank you, thank you. We have- Tarry Singh: Very, very excited. Yaye! Super excited. Kirill Eremenko: Boe definitely added a lot of value to our conference last year, and this time we've got 400 people coming over. So it'll be really cool- Tarry Singh: Waw. Huge, that's massive man. I mean yeah, it's great. You know? Kirill Eremenko: Yeah. Tarry Singh: You know the reason why I think ... I'll just add something to it Kirill. The reason why I think you, and even guys at MIT, guys like Andre Karpathy, who's right now at Tesla ... All these people are ... I think it's important to create this ecosystem with the community, and continue to work the community. We stay away from all these world summits and all these CogX, this X, and that X ... No offense but there's so much air, so much hot balloons flying around. I think the real work is done when you're walking on the floor and talking with ... In fact, I know every person that is .... The community member that walks into all the conference that I've been, the ones which I like to go to, like yours, is they're all walking around with a problem. They're asking questions, they have notes written, I wanna be there. I hate to go to conferences, and that's why we stopped totally. We said we don't wanna be near the World of AI, or World Summit AI, where some business leaders are hanging around sharing presentations. I think, community building is probably the best thing that is there in this. And I hope you keep doing this. Kirill Eremenko: Yeah. I can't wait for you to come, because at our event, we really focus on the, what's it called, inner drive of people. These personal relations, for instance, at some point we just all stand up and we have a dance crew and we're all dancing, jumping, and then after that everybody- Tarry Singh: Nice. Kirill Eremenko: ... you see you get five hugs, five high fives, and really builds these connections between people. After literally two hours after the events start, you can't recognize .... everybody is so friendly with each other. I love it. I love how everybody gets connected very quickly. So that's [crosstalk 00:51:56]. Tarry Singh: Amazing. Yeah. I think ... I really look forward to this. Amazing. Thank you so much. You really got me excited. Kirill Eremenko: Thank you, thank you very much. Okay, I guess we're coming close to the wrap up. Time flies, this is amazing. I just want ... I had this one question while you were explaining the infographic of climbing the mountain of deep learning. If you don't mind, if you have a few minutes, how would a person know ... You've got these five plateaus, which I think ar every descriptive, so first one is statistics-mathematics programming, second one data analysis visualization skills, third is machine learning, fourth is deep learning, and fifth is applied artificial intelligence. So, the question would be, how would somebody know when they are good with the plateau that they're on? When they're confident and that they're ready to move on to the next one? Because sometimes I find it's very easy to be like, "Oh, okay, so I did some ... you know I learned some stats programming. I'm really excited about machine learning, I want to move on forward." And they move on forward, but because they lack that necessary grounding in the ... whether it's stats, or whether it's the programming part of things, they can get very discouraged when they get to the plateau of machine learning, because it's exciting and you can apply it, and they dabble and they get some good results. But because they're neglecting going back and refreshing, as you said correctly, going back and up-scaling yourself in the previous plateau as well, they neglect that part. And they feel discouraged, and they feel like it's not for them, it's not the right thing, when it's really not the case. Tarry Singh: Yeah. I think it's ... And you've trained hundreds of thousands of people yourselves, so I'm sure you must have got so many questions like these. But my personal experience is that, yes, it's very hard to keep a track of all the plateaus when you're climbing the summit. So I say that you don't become and expert if you have climbed the Mt. Everest the first time. Because the first time, you take all the aids, and you're there and you come back, because there's a lot of hand holding going on, there's a lot of ropes. I'm not a hill-climber by the way, but I've heard from people who have done this, some good friends. And then you start pushing the limits and start going without oxygen, right? Many people have done that already, it's proven that it's possible. And so it's almost like building your fitness function, if you will. It's kind of an auto-learn function in which you should intuitively be able to go back. So my advice to people who are, let's say, half way in machine learning, and just thinks ... and even I have that by the way. So it's very normal. First of all, one, it's very normal. If I had to go and look back at the icing formula from physics, and how it applies to the activation function, I have to go back, and sometimes I write them down on a piece of paper myself. Because hey, I mean come on, 24 hours a day, if I'm sitting six, or seven, or sometimes even twelve hour flight from Amsterdam to China, I can do that. I have my laptop, I have all those books in my repository, all of them. So I write it down, and then it helps me. Of course there is a cognitive capacity beyond which you get tired. So I would say, just be selective. Don't worry about an area which you haven't explored yet. You don't have to explain to yourself that you don't understand it. That's okay, you can always come back to it later. Keep almost like little flags, like wait points, you say, "I will visit them later." And going back to the step one is probably the most important, which I realize from my experience, that statistical mechanics and getting deeper and deeper into statistical understanding needs to know how and why it is that way. And then going back into the other ... sort of jumping back from plateau one to plateau four would become easier. So sometimes you have to make big leaps. And sometimes you have to just step down a bit, and take a look at it. The other thing which I want to say is also what I've observed in many people, it's also okay to be at plateau two, for example. A lot of people say, "Hey listen, I understand what it is. I don't need to be in [inaudible 00:56:20] or an expert. Actually I'm great at visualization." And they start exploring like 20 different types of visualization libraries, just as an example. For example, if you're ... and I work with so many bioinformaticians, and molecular biologists, and cardiologists, and pathologists right now. We start looking at the visualization libraries, so all we do is ... The basic stuff is, you show map outlay, you have seabourn, you have cuff links, you have a couple of other stuff. Now I have started seeing that people are looking into the visualization toolkits and libraries that apply to cardiologists that are massive big guys, the VTK and MayaVi, and several other of those which are so sophisticated that they start becoming experts in ... sort of visualization experts. So you're basically a data scientist with that specialization. Sometimes it's okay to also be comfortable with what you're comfortable with. There's no need to start climbing ... not everybody has to do this whole pass of these five plateaus. So choose which really appeals to you, because eventually you will shine there, believe me. Because you will find more things than anybody else would have found. There are many avenues which we ... I'm ignoring them. Although, I jumped into the 3D-ling cancer because I explored, explored, and found more stuff. But then I left it. I hope that someone else takes it and starts [inaudible 00:57:50]. I know you showed something to me six months ago. I improved on it, and take a look what I made. And I've seen one or two people do that, and its was amazing. So it's a two way process. You don't have to boil the whole ocean. Feel comfortable to go down and pick up some more cherries from what you may have picked or may not have picked. But also be comfortable and find with where you are, if this is what you're good at. So I would say don't worry too much about it really. A lot of young people especially get very nervous and anxious, because they think the whole thing about this learning is part of my Masters and my PhD, and beyond that I will not be able to learn. I've been I this industry longer, and you know yourself Kirill, we keep learning everyday so we should [crosstalk 00:58:43] Kirill Eremenko: For sure. Tarry Singh: It's very hard to tell young people that, "Don't worry, it'll come later." They're like, "AH no, I want it now." Kirill Eremenko: Yeah. Tarry Singh: "I want everything packed into my brain right now." But I say, "Well then, you have to wait til the algorithm which in the matrix movie..." Remember in the movie Trinity had to quickly learn how to fly the helicopter. I said, "Wait, we're still working on that algorithm." Kirill Eremenko: Yeah. Elon Musk with his Neuralink. Tarry Singh: Yeah, for instance, yeah. That is an interesting area as well. Sort of patching up the brain and getting ... Literally patching it up, and some `firmware update to get all that information in your head. Kirill Eremenko: Yeah. Crazy world. Tarry Singh: Yeah. You never know. Kirill Eremenko: Thank you. Thank you very much for that excurse and the additional comments on the infographic. We'll definitely include in the show notes. We're coming to an end, running out of time. Tarry I want to thank you so much for coming on the show. Where can our listeners get in touch with you, contact you, follow you, follow your career, and all these amazing things that you're doing? Tarry Singh: Thank you so much Kirill. Obviously it's an honor to be with you. You're one of the shining beacons in this industry. Kirill Eremenko: Thank you, thank you. Tarry Singh: It is true. I've followed some of your trainings myself. I continue to talk about your trainings in every class that I teach. On LinkedIn I'm there, quite active. Fortunately, LinkedIn's algorithms, they've been improved since I think July. Things are working very well for LinkedIn, and also for us. I'm on Quora quite often. Lately again jumping actively to Quora trying to answer questions as well. I think these are the two platforms. I'm also on Twitter, although I occasionally respond only to my Silicon Valley Networks, and all the researchers form U.S. or Canada. But otherwise, I think LinkedIn and Quora are the great place to be, but Twitter is also a place. These are the three places. Kirill Eremenko: Thank you. Tarry Singh: And please reach out to me. I personally like to talk to people. There's a lot of time that goes in to do that. But send me a note, you will always get a response. I'm just a normal guy, I'm not sitting on a high-horse, some elite researcher who works at Facebook or something. Honestly, I just like to talk to you guys. So please, let's just be normal human beings and have fun learning- Kirill Eremenko: That's so cool. Tarry Singh: ... deep learning. Kirill Eremenko: That's so lovely, thank you. And I'm just gonna add to that, Tarry has a blog, so Tarry Singh, T-A-R-R-Y Some very interesting topics are discussed there. And of course, if you don't mind me sharing,, Something exciting is going on there. You've got a countdown timer, what's that all about? 18 days, 16 hours, 16 minutes, and 22 seconds. Tarry Singh: Yeah, yeah. So the platform, Oh my God. The platform in which we have like a hundred and fifty applicants, or probably a lot more. I know there are thousands, I looked at the landing page, it was like, "Oh my God." So people wanna go for research, they're applying for research and philanthropy. I forgot to mention, maybe it's worthwhile mentioning- Kirill Eremenko: Sure, of course. Tarry Singh: ... that we are going to be collaborating with Hult Foundation, that's H-U-L-T. And Hult Foundation is ... so Bill Clinton and Hilary Clinton are also ... they contribute to the foundation, in fact, they also inject cash in it. Hult is a very wealthy Swedish family, who also set up the Hult Business School a long time ago in U.S. So we're going to be collaborating with them, and I will be in London I think sometime in August to be coaching and mentoring. It's a week long program, 40 start-ups in AI. Hult is something which you will see more announcements coming as we try to launch a prize. Something like you mentioned Peter Diamandis'. But I have an idea how to do it, sort of ... we are calling it a deep [inaudible 01:03:02] shield. So getting more and more people into AI. So that is a philanthropy initiative. So there's a lot of stuff coming Kirill. There's a lot of work. I know people are really upset at me that I put so many things in parallel, so everybody has to work. But yeah, we are full of energy. And we'll stay healthy. So I hope we can achieve the mission that I think deserves achieving. But I'm looking at you, I mean you've done it. So I think it should be possible for me to do. Kirill Eremenko: And I'm looking at you, and I'm like "Yeah, If you..."- Tarry Singh: You learn from me too. Kirill Eremenko: Yeah. For sure, for sure. And I'm sure it's all going to go well. Can't wait for the timer to hit zero and for you guys to unleash the power of AI to solve worlds problems, and bring the word out there [crosstalk 01:03:49]. Tarry Singh: I hope so. The developers are working hard for the platform. Let's hope it... Thanks once again Kirill. Amazing, thank you very much. This is the first time we've spoken, but I have followed your work for such a few number of years. Great job for doing this. Without you guys ... every interaction leads to some expansion and idea, and thought. And thank you for this great talk. My day is now full of talking to another gentleman who's done this for so long. It's a thankful day, it's a grateful day for me. Kirill Eremenko: Thank you. Thank you so much for coming on the show, and I had a wonderful conversation, and I'm sure our listeners enjoyed all the insights you shared, and all the inspiration that you just convey with your energy. I can feel it from over here even though we're on opposite hemispheres in the world right now. Tarry Singh: Amazing. I'll talk to you soon, and see you in San Diego. Kirill Eremenko: See you in San Diego. So there you have it my friends. That was Tarry Singh, founder, executive, philanthropist, researcher, and as you'll probably agree with me after today's session, just a very, very nice guy who gives back so much to the data science community. And my personal favorite part of today's episode was when Tarry described the three different components of his strategy, when he talked about the enterprise advisor, the research arm of his business, and the philanthropy component. So when you put all those three together, it becomes pretty clear on how he has been able to make such an impact on the world, and empower so many individuals in becoming data scientists, becoming more data literate, becoming data advocates. And I'm sure he's going to continue this mission going forward. If you'd like to get in touch with Tarry or follow him and his career then make sure to check his LinkedIn and Twitter, we'll be sharing those in the show notes at We also mentioned quite a few of his websites and different undertakings. Those links will also be available there. And as Tarry mentioned, he'll be coming to Data Science Go 2018, October 12th, 13th, 14th. If you haven't gotten your tickets yet, you can get them at We've still got the early bird prices available. They're increasing this week, so make sure to jump on board and you'll see there're plenty of wonderful, amazing speakers, just like Tarry. We've got twenty speakers coming to Data Science Go, and we can't wait to see you there. Hope to see you in October, and until then, happy analyzing.

Kirill Eremenko
Kirill Eremenko

I’m a Data Scientist and Entrepreneur. I also teach Data Science Online and host the SDS podcast where I interview some of the most inspiring Data Scientists from all around the world. I am passionate about bringing Data Science and Analytics to the world!

What are you waiting for?


as seen on: