Business at the Edge Episode 5 - Angelo Kastroulis

Business at the Edge Episode 5 - Angelo Kastroulis

Show Video

hello everyone and welcome to business at the edge our video interview series with thought leaders and practitioners at the intersection of edge computing ai iot and other innovative technologies my name is blaine matthew and i'm ceo of protecso the edge computing platform as a service now let's get started joining me today is angelo castrulis from ballista a consulting company at next-gen vc and he's going to tell us more about that and of course angelo is also well known for his counting sand podcast thank you for the time angelo we're gonna have some fun today thanks blaine thanks for having me excellent fight we really do appreciate it so why don't you start by telling us a little bit more about your personal backstory before we find out more about ballista how did you get to where you are today i've been listening to the counting sand podcast and so heard some of the story actually but very interesting because obviously you're a technologist at heart but you really make the connection between technology society science philosophy many different areas and as i'm listening to it i'm almost reminded of the the old dos equis commercial you know the most interesting man in the world well you're even look sort of like the dos equis guy i would say but you've got an amazing background so tell us how did you get to where you are today well well i don't know if i could live up to that that's you know you're right i have been in technology for many years and uh i think coming from the trenches like that really gives you a certain perspective um and it wasn't until later on in life that i realized that that maybe 20 years into computer science i don't know 20 or 30 even maybe you get into and you realize that that there is a purpose behind all of this information that we're doing these computations all these different things so i went back to grad school to try to make sense of it try to figure out this new thing called machine learning and ai and try to see how can we analyze this for meaningful ways of doing it i got into the harvard data systems lab and that is when i i think it kind of opened up to me a little bit where i saw the world a little bit differently i got to see the insides of a lot of these commercial databases and how they works data systems i should say and you start to really see that these are not just very huge technical problems but there's a reason you're doing them and so i think for me uh what kind of brought these different things together you know the technology side and it's almost like a venn diagram and that's what we say about data scientists right they're part mathematician they're part computer scientists they're part something else some kind of domain that you're interested in yes and i think that also applies to just this idea of data systems moving data along so that's kind of what gravitated me and i realized that there's just not a lot of people doing this and so it really appealed to me and so what what spurred you to actually form ballista and maybe you can tell us more about what ballista actually is yeah sure um well it's one of those things i think where where you see a need in the market you you see that computation is important to a lot of people big data and some of those technologies were starting to really come to fruition when i was in grad school i i had my thesis was actually on my research was also based on using ai inside of a database to try to make data systems faster and and so you know i got my that was my thesis i ended up getting a patent and it kind of made me start thinking of things differently and then you brought it back to consulting i was an independent for a long time and brought it back to consulting when you realize what you wanted to be when you grow up and and people will say well how this we can't solve this there's no system out there and you say well i can solve this right let's build our own and i think that that's kind of what spawned to build you know well what is ballista now i met three awesome partners and we decided to kind of put our forces together and uh inform a company called ballista and what does ballista do again describe the kinds of projects you work on for clients yeah so ballista tries to tackle the hardest problems in computer science and so we're a consulting firm park consulting firm in part vc and explain what that part means but on the consulting side we're we're software engineers but we're also data scientists mathematicians and and we understand domains and so we have a lot of experience in healthcare and iot in fintech and so we're trying to kind of bring all the things that we're learning from these different areas together for example one of my biggest customers was wanted to build a clinical decision support system and so we brought our expertise and data systems together to build something to compute quickly and and it's about massive scale quickly it's not just about doing something fast and running on a million machines yes but can you do it on 100 machines right that's a whole different problem so that's kind of how we uh we've we've kind of always evolved now the other side of it is i think consulting services is important it's got to be part of your dna i think you and i were talking about this in the past that think of life of a computer system or of a business as a genetic algorithm right and you're gonna do the same things you always do generation after generation but unless you insert some randomness every once in a while you go do some fintech work or do some retail work or do some other things you're not gonna inject any new knowledge into what you're doing that's why we dive a little bit into things like quantum computing which is most people be thinking that's crazy that's not what you do but you have to dive into it because the world is going to eventually evolve and you need to be aware and and bring those kinds of solutions together so that's kind of what we do we try to always be on the bleeding edge and you know bringing in even cutting edge research research papers from the universities stuff hot off the press and trying to implement it wow we could go in so many different directions right now but it sounds to me like uh the kind of projects you get involved with with clients are the hard ones the complex ones the ones where the answer is non-obvious you know it's not just oh let's bring a few developers together we we know exactly what to do it's just a matter of of time and how many bodies you can put on it instead you're solving some of the more complex challenges and with technology and computer science today can you give us a couple or an example maybe of a hard problem you guys have tackled in the past yeah actually this is one we're working on uh currently it's perfect it's really interesting so so one problem is that the the federal government is very interested in reducing the cost of health care right and there are many ways to go about that um one of the ways is just trying to understand if we take some guidelines how are they followed what what is the the outcome and do they affect outcomes for positive but there also is a cost associated with them and does it affect cost so that requires computation on massive scale right these payers have hundreds of millions of patient lives that they're insuring and so to be able to report this to the federal government there's a lot of data that has to always be crunched i mean you talk about billions of pieces of data yes um so we we started tackling this problem with a few customers and looking at it in the sense of you know the other solutions out there they were looking at like well this takes you know eight hours to execute this computation and so we said okay shouldn't take eight hours what should it take i said five minutes so we built systems that actually compute in 20 seconds things that just everybody thinks this is impossible that's because we're not applying the most cutting edge thought process we're trying to kind of do it in a conventional way and that just doesn't work right right right so i think that's a proof right yeah it's not just about applying brute force to a project but applying intelligence to it is it about the application literally of applying ai and machine learning to solve these these challenges is that generally the solution or not always not always it's it's about i think looking at the value chain and then saying well if i change things a little bit right maybe like if i look at it holistically and i rework the way we think about it ai has a place you know we've we definitely have expertise in ai in optimization which we've proven that you it can be the human optimizing any day and you can watch it all the time so yes that certainly has its place but sometimes nobody asks the question do i even need that like why does that even exist and can we do this a completely different way that eliminates these links in the chain so that our value we're more direct to the value right the chain is shorter and we can get directly to the value really interesting i want to circle back to the vc thing after i don't want to lose lose that that you brought up but i want to keep our current thread going for another few minutes so what really excites you about what's going on in technology these days what is uh what do you think are the next or the current and and maybe next interesting uh areas of development advancement we're gonna see i think there there are a lot of interesting things coming um i think ai thinking about how ai is being used i think it's you know it's still kind of in the hype cycle a little bit and i think a lot of organizations don't know what it is and you see a lot of companies throw ai out there like we do all this ai and it's not uh you'll see that yep but i think we're we're maybe coming to the point where we're realizing that it's not all about moonshots that we can use ai to get really huge gains by just eliminating some some small problems so there's low risk to eliminating them but high reward i think that is something that excites me um about this this is kind of the thing that we all it's intuition that you'd say well yeah of course that's what we should have done but we always fixated on the moon shots so i think that that's an interesting uh wait place that this could go i i do feel edge computing is going to be revolutionary in the way that the world views their problems it is one of those things when i talk about changing the value chain and thinking about it in terms of why not why are we even doing that how can we eliminate this problem and i'll give you an example for example if you're if you're trying to compute data and this is kind of how we have seen the cloud forever throw as much as you can up to the cloud and then we will compute throw it into a lake we'll compute it later because i don't know what insights i'm going to get well the lake ends up growing bigger and bigger and bigger we create new problems for ourselves we have to clean it and do all kinds of other things and then all we require now is a more powerful cloud and the cloud keeps growing and it it turns on its head what we wanted to accomplish and that was let's reduce costs by putting in the cloud now you go well the cloud keeps getting so big that i'm doing so much work in it right right yeah you know and and i have lots of thoughts on that we talk about that a lot on the podcast this idea of just constantly keep growing it well the question we should be asking is is there just a different way to do it that's how you're going to get exponential gains you're not going to micro optimize your way to a better cloud i think that that is a is a big thing that we're looking at of course quantum computing is the other thing that's been on my mind a lot lately you know how can we again quantum everybody thinks of in terms of moon shots but i think that they're starting to emerge as a technology that we can actually start using for optimization problems now that can have some return so you think uh actual real world applications of quantum computing could be in the next few years not a few decades away i do think so um when you think about and things like annealing they're very good at um optimization right solving for a best path or best lowest cost right they're really good at that in a lot of ways many problems could be reduced to that now the problem with the moonshot way of thinking is not every problem will fit in that space so it becomes very hard to mathematically uh look at a problem in a quantum way but there are a lot of problems that i think we can uh reformulate and reframe now in the future i don't know that we'll need to re-frame all of our problems like that will be it will grow sufficiently that that's it okay that's the 10-year horizon but i think today we could start seeing some problems that would definitely benefit so you brought up ai machine learning edge computing what about iot i often see those as being back to the venn diagram being a way that these three things are very close can be or must be very closely related to each other thoughts on the evolution and current phase and future of iot technologies in general yeah i'm glad you mentioned iot it's it's one of those things that um has found itself into so many areas that didn't exist before um i'll give you an example wearable devices for fitness for example you know everybody always thought well my watch is capturing my blood pressure and it's got my heart rate why do i need to go get another device or go to a doctor to make an appointment and it was there because a doctor is responsible for the reading he can't trust your watch he doesn't know that you're using it right or that it's on you properly or whatever kovid changed it so now we were by the necessity were forced to look at these devices as first-class citizens so that's just an example of how i see iot now it's it's being elevated to the point where people say wow these things are really useful and everything was okay we relied on it and everything was still okay and it provides so much rich data that i can actually it can be actionable for me and i think that that's what we're seeing now that's in healthcare who has been i think very resistant right is that fda approved we have you can't use that device but they were very resistant and now they're becoming much more open to it granted out of necessity but it's proving a point i think other industries which leads to the reason why i believe you have to be involved in various industries other industries have been on the understanding the value of iot for a really long time you know so they were using these devices to try to understand what's going on at the edge and so bringing that knowledge over to other industries i think you're going to see an explosion of iot which leads itself back to my original problem of that's going to generate more data and if we throw it up to the cloud what have you just done to the cloud yep yeah exactly very very interesting perspective and i yeah i couldn't agree more so let's change gears tell me a little bit about the vc sides here obviously not only uh consulting with companies helping to use next-gen technologies to create interesting solutions uh non-trivial solutions to complex problems but it sounds like you're doing some investing in this area as well yeah so i i would say the best way to describe it is that technology cannot live in a vacuum so you can't think of it in terms of um an organization needs to modernize so they'll come to us and say how do i modernize my stack the real the real gain is not just from i have this project i've already designed it or can you help me design it this is what i want to do the real value is one level up where we say this is the business objective i'm trying to achieve and then you evolve in someone in that conversation so they can say well let's re-examine your value chain and that that means the technology is there's an investment in it and i think people understand when they when they go to modernization they're thinking i have these aging systems i need to replace them the one piece of advice i'd have is don't think of the world like that you it's a constant investment you have to make because if you're going to let the system age and then reinvest you're going to be so afraid about what it's going to break that you're going to try to lift it and fit it and all you've done is displace your problem yeah so i think that you can't think about technology in a vacuum you have to have a sense of business process you have to have a funding to do it um and then and then of course you also have to think about the sales side of it right how do you partner who should i partner with to be able to deliver my product so i think that there are those aspects that that we have some unique skills in that area and that's why i thought this makes a lot of sense for us to be involved in the other side of it is if i were an investor and someone was asking me for technology you know money for technology maybe i'm a little biased because i'm technical but i would look at it and say do i have the confidence that you can pull it off and it's so much easier to get some money and then say well the money comes with we're all still going to bring the expertise and we're going to do it right and because i have a stake in it i won't want to do it right and i want to do it as quick as possible to get you to market get out and so that's you know we don't we're not here to be part of your your you know your journey forever we want you to get on your feet and get going that's different than than just purely consulting right totally so since some of this audience likely will be startup ceos what kind of companies do you think should be maybe approaching you what's what's at a high level what are your criteria for what you're looking for yeah i think yeah so yeah it's hard to say that the criteria we're looking for i can tell you what we found to be successful um organizations to have started in and have some revenue and have figured out that that they they've they found their area in the market right you're not necessarily coming to us and saying what i got this great idea is there a market for it i think the better approach is to say yeah we have this idea however these are the things the problems we're seeing with with what we're running into as an organization and it might be we need some funding to grow the next level we might we need some partnership advice or business strategy or find a ceo but it could also be we need to modernize the stack because we can't get to the next level of growth we can't cross the chasm if you're in that area trying to cross the chasm that's probably the best fit makes perfect sense given your background and focus and all the stuff we talked about here the consulting side of your business so perfect uh maybe tell us a little bit more about the counting sand podcast what was the thought process behind it as i mentioned earlier i've listened to most of them now really really different take on technology and and related areas so what's what's going on there you know this is it's it's really um it's been a great journey first of all but you know what made me decide to even do a podcast um you know i went through this maybe everybody does this i don't know i hope every leader does this in any organization sit down and think about do some exercises to determine your core values and when you do think about what you're good at and what you're not good at in terms of those things and it's funny i sat down and i was doing this exercise and then i and then i stepped away from my life for a moment and said what should i be doing if this given this and i chuckled and i said i should be a podcaster and i said but that's ridiculous and i walked away and then about a month later i thought why is it ridiculous i shouldn't be thinking like that and and so the idea i started doing some tests asked some friends and and kind of tried to put together the right uh format i think the thing that i really love about the podcast is you'll never hear me pitch my company in it i in fact i only mention it at the end when i say you can follow me on whatever i i don't because it's not about that it's about giving a little bit of knowledge back to the community and so i think that's what makes it kind of cool it's this mix between research and application and i'm trying to kind of get everyone to understand if you don't be afraid of research it's not crazy there's there's the giants have done some great things you can build stuff on their backs and then we can we can apply it it can be done and i think that's basically the thing we're trying to kind of get get going well i think you're accomplishing it and i highly recommend the folks that listen to this to check it out if they're not already uh doing it because uh i've got a lot out of it in in the last few months since i've been listening thank you for doing that so uh maybe to wrap it up yeah thank you maybe to wrap it up so any uh predictions i guess or thoughts for 2022 and and beyond you've already shared a few but uh what have you got yeah yeah i i think if you don't mind i'm going to rehash a little bit or go back on in a little more depth of the one you talked about that we talked about i uh edge computing there's there's a few drivers here that have been happening for a few decades and they're catching up with us and i think especially this year when you think about uh for the longest time compute has been one of those things that you know moore's law and we know it's coming to an end why because we're down now they're talking about the one nanometer process right i mean you can't go to zero so there's a one nanometer process that also means that clock speeds have reached their max so what do we do we put more cores on a die and then what do we do we put more machines in a data center there's only so many machines you're going to be able to pack right we can't make it any smaller so there's only so many machines you can pack into these data centers so you have to think about the world a little differently and i think if we the other implication here is when you're when you're growing this data center and i think we'll talk about this on a podcast i'm sure but as you grow this data center you're using more and more energy it's got uh it's more and more cost right what happens is the cost per cpu now is reversing and we got to stop that trend so yes quantum shifts will change the world in the future but now you have to think about how do we just get eliminate that problem why don't we just not compute inside the big cloud i don't mean entirely i just mean can we push that closer to the source so that the source is giving us a little cleaner data and when you think about the source that means the edge what's the edge it means everything it means your phone it means a boat it means an airplane it means a iot device a windmill it means that a windmill yes exactly so if you can push computations down there you've now you've created orders of magnitude improvement in the cloud and that's what i mean those are those are not those are you know to get a qualitative change you have to rethink the value chain to get a quantitative change you can micro optimize this is not a micro optimization problem we we need to take it qualitatively to another level agreed obviously i think that's going to be yeah 22. i think 22 is going to be a big year for that okay yeah i believe so too and it's interesting you brought this up because we didn't uh talk about this beforehand i didn't know what you were going to say but but obviously i i fully concur with your assessment and i think right now uh we've only just begun scratching the surface of which what's possible with processing data closer to the edge all the way to the far edge and the device edge we're just just beginning to do that and it will be you know i think a pretty radical in a positive way transformation to how businesses are able to use compute power no doubt about it i'm going to add one more thing to that blaine i think that i think you're right and what you said kind of reminded me of one of the things i care about ai right um i think edge computing we couldn't have done this because a not logical question would be why didn't we do this 10 years ago well we couldn't have done it because we were not at the point where the cloud was a problem right and we weren't at the point where we could have compute feasibly be inside like an iphone is more powerful than the queue the land of the apollo 11. right you can't have you didn't have that but now what is really interesting is the next generation of issues is going to be we have lots and lots of edge nodes what do you do how do you optimize them now ai is going to be involved in my opinion and then we say here's how you optimize it you optimize it by by making a.i handle the millions and billions of nodes that are out there yeah intelligent optimization of that edge network fundamentally right okay well i think if there's anybody that has the ability in the team to do that i think it's you you and your team so let's let's make it happen i'd love to help you on that absolutely thanks plan i really appreciate that you bet well i think that wraps it uh again angelo thank you so much for joining us today it was a great conversation as i knew it would be of course i probably i'm happy to be here thank you for having me yep and for those that are interested in hearing more of angelo's thoughts on these and similar topics you can connect with him on linkedin of course and on twitter twitter at angelocastor kstr and also check out ballista.com

two else can't can't miss it and of course you can reach out to me anytime on linkedin or at edge protecso.com thanks again angelo my pleasure

2022-05-13 22:44

Show Video

Other news