george grombacher 0:00
Come on one Leffler This is George G. And the time is right, welcome. Today’s guest is strong and powerful. Luke Ericone. Luke, are you ready to do this? Let’s go, let’s let’s go. Luke is the CEO and principal data scientist with air core AI. They’re helping companies build state of the art machine learning and artificial intelligence programs. Luke, tell us a lil bit about your personal life more about your work and why you do what you do. I am

Luke Arrigoni 0:37
married with three kids in Seattle. And I build machine learning programs. And I’ve done this for a swath of companies, from startups, to very large companies. And I think I really enjoy kind of the the state of the art and being able to bring that to different groups these days.

george grombacher 0:57
Nice. So I was I was perusing your LinkedIn profile. And I saw that you were working on blockchain way before the rest of the world realized that blockchain was a thug.

Luke Arrigoni 1:09
Yeah, that was it was kind of fascinating. There’s been so many kind of ebbs and flows with Blockchain. It’s been something that’s been, you know, largely discarded a decade ago to a few years ago, there’s a huge spike. And then it was forgotten again. And then in the last year, it’s just big again. And I think, several years ago, when there was that huge uplift, it was great being able to be one of the development teams on that.

george grombacher 1:35
Yeah. Interesting. So when I hear machine learning and artificial, artificial intelligence, I think I have no idea what that means. It can mean just giant, it can mean. So what, what does that mean,

Luke Arrigoni 1:51
you know, you’re not far off. I think when most people hear AI, ml, they think, a spaceship going up somewhere a robot folding your laundry. But the reality is, it’s actually something far more practical, which is people will actually build machine learning to do a very simple task, right? It might be being able to categorize a package that’s going out, or it might be able to say, you know, what kind of thing is in this image, it’s actually things that are quite mundane and boring. But when you bring them together, you can automate lots of laborious tasks.

george grombacher 2:26
Okay, so it’s a function of, if I can conceive it, it can be built kind of a thing.

Luke Arrigoni 2:34
Yeah, you know, I think the the only rules around it around data, right? If you, if I don’t have data, I can’t do anything. Because we teach these machines. We call it machine learning, you know, colloquially, it sounds like we’re sitting down and talking to a machine, in a way we are, what we’re saying is we say, here’s the data, here are all these examples of something that I want you to predict. And then we use math to help that machine get to that place. Without that data, we don’t have a language to teach these machines. So essentially, if you have a whole treasure trove of data on something, and you want to do a prediction within that data, yeah, you can sit down and build something kind of cool.

george grombacher 3:13
Nice. So we need the input. And I imagine that there’s probably organizations out there that are drowning in this information, though, they’re like, Okay, I’ve got all this info. Now what?

Luke Arrigoni 3:28
Yeah, your intuition is spot on. There are companies that have this treasure trove we talked about, but they don’t really know what to do with it, right? It’s kind of sitting out there. In the void, maybe it’s a little dirty, and it’s kind of scattered. And they, they know that there’s there’s a solute system somewhere that’s collecting it. But at the end of the day, they need to figure out if they have that predictive value. And so what it can do is they can hire a company to do what we call discovery, where you say, You know what, I have all this data, it’s out there, it’s ready to go. But you know, I need to figure this out. Well, you would hire a firm that would go through that and say, I kind of have an idea. And that’s the first step for a lot of these large companies, or even small companies with huge amounts of data.

george grombacher 4:12
So there’s organizations that actually will look at everything and say, here’s sort of what you got. Here’s maybe what you think. But you don’t?

Luke Arrigoni 4:22
Yeah, I mean, of course, selflessly. I’ll say that, you know, that’s what we do and their core, but there are there are other companies out there as well. Largely, you’ll find it with AI ml consulting. This will be one of the first steps on a standardized process almost these days.

george grombacher 4:38
Okay, well, let’s just walk through that process. Yeah.

Luke Arrigoni 4:41
So you basically have a lot of data and oftentimes, you’re in a mid level managerial position, and you’ve been basically given a budget to do something cool. And I’m not joking when I say that. Oftentimes, they’re given the the command to do something cool with the data. And that’s kind of the state Right, that’s That’s all right. And so, you know, there might be discovery that you do on your team. But ultimately, you would reach out to a company like mine. And you would say, hey, what do you think is possible? And we’d look through the data, we look at your business, we’d ask about what your customers are concerned with, or pain points, we really try to understand the fundamentals of your business, because we know you don’t care about the math, right? And then once we have that firm grip on your business, then we can say, Okay, we know that your customers are struggling with blank, did you know that your data contains the answer for that? Why don’t we build an ML model, and we connect the dots, and we wow your customers?

george grombacher 5:36
So what are some of those are? Are there through lines? Or does it always depend? When we know your customers are blank? What is blank a lot of the time.

Luke Arrigoni 5:47
So like, here’s a good example. Sometimes, if you’re an E commerce shop, you have this kind of window shopping experience, where people may be like your brand, but they don’t necessarily know what to buy from you. This happens when you have you know, kind of a very specific product that you sell, or a specific field or a brand that you’ve cultivated with your following. And what we can do is say, you know, this person is like 1000 other people that came before that really enjoyed X product, right? Whatever that product may be, maybe you know, it’s a coffee mug, right? But it’s a coffee mug that has your cool emblems on it. And so what we can say is they go window shopping, we could detect with machinery and say, Hey, this is now a window shopper in real time using machine learning. And then we could say, Okay, this window shoppers like this cohort of people that love the coffee mug, so when they’re browsing through, let’s put the coffee mug at the top. And now this is all very seamless. Like I said, it’s a it’s an orchestration of little tiny, boring things to make something cool. So now I’m a customer on your site. And suddenly out of nowhere, like magic, the thing I’ve always wanted now appears, it’s there’s no fanfare to it, it’s just right there. And a lot of times companies like Amazon and have these things built. But now small companies can build these same things using machine learning.

george grombacher 7:05
Amazing. Alright, so when we talk about companies have treasure trove of data? Is it data or data? Look, I’ve heard

Luke Arrigoni 7:15
it both ways. I really, I’ve never, I’ll never correct anyone on it. Because even I don’t know. And I’ve heard it said, the different ways for the last 15 years. So okay, now you might even see us two different versions in the same conversation. So

george grombacher 7:29
I appreciate that. Where so so this mid level manager has a budget that come to you and they say, Hey, I’ve been told to do something cool. Where is all of this data living?

Luke Arrigoni 7:41
So great question. A lot of times it’s stored in, in a database that’s used for a completely different purpose. And that by itself is challenging, right, you’re like, I can’t go and touch the thing that is running our store or running a warehouse, right? Like, you can’t mess with that, we have to pull it out first into something safe. But oftentimes, it’s an output of transactional data. That’s the term we use for when you go to a store and you buy something and there’s a card and it gets saved. And we that all that gets shoved into a database somewhere, and no one knows what to do with it. All that transactional database, or data from the database is typically what we look at first.

george grombacher 8:18
Got it? All right. And so you can you can safely extract this information or locate it. And you say, Okay, well, it turns out, you have information about what how long people are spending, what they’re actually looking for, really, this, I mean, it’s your job to sort of interpret this language of numbers and stuff like that, and help people understand in an easy to understand and digestible format, and then make decisions about things.

Luke Arrigoni 8:49
Absolutely, it really is our job to make it so that we are a there for your business. And you don’t even think about the math, we don’t want to be magicians, right? We want to be able to explain what we’re doing. But at the same time, we don’t want to bog down the process of you and your customers by saying, Here’s a math lesson in linear algebra. So we want to be able to remove that part so we can all have fun and build something cool together.

george grombacher 9:13
Yeah. So do you come back in I imagine you do your best to figure out how people like to make decisions about things and say, sort of here’s a download of what we have. Here’s some different thoughts and observations on what you potentially could do. And

Luke Arrigoni 9:30
yeah, spot on. We out of discovery comes this idea a project is born. And the project should be at a place where you can take it to one of my competitors, like it should be a fully formed thing, right? And it should be, this is the data we have. This is maybe even a prototype we built or designs around what we have, and we think that we can get to this place. Is this a place you think would have value in your business? And if you can come to a place with your budget and with your time your reasons So yeah, this makes sense, then we’re on, we have a fun project ahead of us.

george grombacher 10:05
Got it. Okay. Nice. And how, how intense is that on your end from a human being working? Like how? How is it actually being made on on your end? If that’s an approach?

Luke Arrigoni 10:23
Yeah, no, we can go, let’s dive a little bit deeper into it then. So when we get all the data through, of course, lots of NDAs are signed, we make sure that, you know, we’re completely insured. So if you’re afraid we’re going to leak it or anything, you could sue the hell out of us. Like, there’s a lot of the legal side that we’ve thought through over the last decade, to make everyone safe in this experience. So once the data comes to us, we’re basically at a point where we say we know their business, because we’ve interviewed with them a lot around like what they’re concerned about, let’s go diving to try and figure it out. And so let’s talk about the the coffee mug, right? If they say, you know, we have these window shoppers, they leave our site, they never convert, but we know that they want something we just don’t know what we go on, we find a way to categorize. Can we classify who they are? Right? So we look and say, Hey, these people on this DSL fit into this model. And this, these people fit into that model. So let’s build a little model real quickly. And maybe it’s only 80% accurate. That’s not really that great. But it’s enough for us to start saying, I think this could be 95% if the client had the resources to do so. And so then we’ll mark that and say, Okay, that’s the first project, what’s another problem that they have? And then we’ll go and do it all over again. Basically, we try to get to a place where we think we can build something using the data and the evidence they’ve given us know, initially.

george grombacher 11:41
Nice, okay. And I suppose it’s, it’s all relative, but when when you get back together, and you say, Okay, you have this conversation about what’s possible, and they say, Love, this is amazing. Wow, awesome. We would like to do that. Is there a timeframe? That’s that’s customary? Or does it just depend to actually bring the thing to life?

Luke Arrigoni 12:03
It is very relative, but you know, it, let’s give real numbers just so people have something to go on, right? Oftentimes, you we think about discovery has been about a month, maybe six weeks, we think about a prototyping period has been roughly under six months. So I would say most most clients are comfortable with like a one to three month prototyping and prototyping is you take the discovery here, like you get as you run as far as you can, that 95% accuracy, right. And then the last stage is what we call production icing. And that’s when the prototype does exactly what we think it does. And it does it really well. And everyone is now comfortable, you have other people in your organization sold, and you use us to do that selling, because oftentimes, if we built this, then you need us to go and talk to a different team and say how cool it is. So they get excited, too. So it goes out to prod. And then we make it so that when it is in production, it’s safe. So we’re not recommending, you know, a wrong product to your customers. And then it’s you know, deleterious to your brand, right, we do a lot of that testing, we make sure that it’s stable, so your site doesn’t go down when it tries to access it. And I would say that the longest part is probably the most boring, but it’s making sure that this thing lives seamlessly in your environment. And that can take up to six months as well. So a lot of these projects, you know, who you think three months to about a year,

george grombacher 13:23
got it? Make sure it lives seamlessly in your environment, so that it doesn’t escape and destroy everything. Look,

Luke Arrigoni 13:31
yeah, you know, I’ve seen a lot of these, these products, and people get really excited. And then they pushed about, and then they perform very poorly. Or they’re unstable, because it’s not the normal we call a stack when it’s written in one kind of language, right? So it’s maybe not their normal stack. And, and we of course, don’t do that we try to write everything in the stack that the customer has, but we’ve seen it before, where someone will kind of rush something out, and they’ll put it on a on a server somewhere. And then suddenly, like a system that was working stably before, is now served by this new system that is, you know, crashing every day, right. And so the excitement around ml kind of goes away when everyone else in your organization doesn’t care about it as much, but does care about the fact that they’re getting more customer complaints, right? So when we say seamlessly, we make it so that no one really cares about it, unless you want them to write unless you want to highlight this feature in the newsletter. Look at this cool window shopping feature we built for you. But for most people, at least most of my clients, they really don’t want customers totally aware. They want it to feel like just part of that experience they have within their product.

george grombacher 14:41
Yeah, that’s a great point, right? You don’t want it to be intrusive. You just you don’t want to just you said it better than I can sum it up. So nice. So I listened to a podcast, the Tim Ferriss show and Eric Schmidt from ex Google was on And he was talking about the future of artificial intelligence. And I was like, Oh my gosh, we are in giant trouble as human beings. Yeah.

Luke Arrigoni 15:09
There’s probably I know, everyone wants to bifurcate the world, right? There’s one of two types, but I’m gonna do it right now. Yes, what types of people in AI, right, and you have one camp that says, AI is going to destroy the world, right? It’s going to become sentient. And you have like, people like Elon Musk in this category, right? And they’re, they’re very vocal about how big and important machine learning is going to be and how it’s going to change a lot of our patterns. And then we have a different camp that looks at at that and says, maybe, probably, but definitely not anytime soon. And I’m definitely in that camp. I’m in the camp of people like that probably don’t have a practical knowledge of how far mymail is today, they have kind of an executive knowledge, right. And so I know that sounds funny, saying that Elon may not have, like the most hands on, but we all know that he’s not down in the weeds building these models, right? These machine learning models. But the practical reality is that AI has decades to go before we have to worry about it. And in the meantime, you know, I have a Tesla, I would love for my car to actually drive itself, right. It really doesn’t. And so that that highlights this, we have, you know, some people out there talking about how it’s going to destroy the world. And yet even their own products struggle to make good on the very simple ml AI promises they’ve made. I hope I’m not too critical of anyone. But uh, I just I hope when people hear those kinds of like, glorious headlines, the reality is like everyone struggles with it. Even Elon struggles with AI right now.

george grombacher 16:47
Yeah, it’s fascinating. So I appreciate I appreciate that perspective. Did you say that on one side? People say that, yes, it is going to eat the world? And then the other side? You said? Maybe probably it is?

Luke Arrigoni 17:02
Yes. I think it’s inevitable. There’s a lot of things in our in our reality that are inevitable, right? That’s basically have to say, when do we make decisions on it right? And like, so everyone in the world is struggling with global warming, right? If everyone in the world was aware that it would happen in the next six months, there’d be massive changes, right? We’d all like never drive a car again. Right. But we don’t know if it’ll happen in the next 600 years or the next six months. We just have estimations. Same thing with AI, right. It’s the timeline that matters. Like for instance, if I told you that Google was going to go bankrupt, that would be worthless knowledge, unless you knew which quarter which financial year they were going to go bankrupt, right? You’d either say, Well, I guess I’ll sell the stocks now. And you know, that’s medium. But if you knew that in q4, or of 2023, you would short you buy all the cheap shorts right now on Google. So timing matters, right. And I think it is inevitable that AI will eat a lot of our a lot of our tasks, a lot of our labor. sentience is a question mark, I think for most people, but um, it definitely is going to wow, and scare people in the future. But how far that future is from today? Is something I think that is greatly debated.

george grombacher 18:16
Yeah, I appreciate that. Like it. Well, Luke, the people are ready for your difference making tip. What do you have for them?

Luke Arrigoni 18:24
Yeah, so I’m often tasked with building really complex things or designing complex things. And I used to really struggle with sitting down and being like an eight hour block of trying to design that never really happened. And so I started chunking off parts into my calendar. So if I had a really huge task, I just think, you know, what are the five things that make up this task, and that actually scheduled things as ridiculous as like lunch, or a 15 minute break. And so when I sat down to do deep work, I have almost a calendar list that I’m going through and like the appointments are coming through my calendar saying, this task should be about this done. And when I’m behind, I can start adjusting that. And so I guess that the big tip is, if you have a huge thing in front of you, don’t try to eat it all at once. Definitely try to break it down into smaller parts and use your calendar as a good system for that has been a very effective tool for me.

george grombacher 19:18
I think that that is great stuff. It definitely gets can eat an elephant one bite at a time. So that is music to my ears right there and actually scheduling things in it’s such a simple, but massively powerful thing so well. So Eric, thank you so much for coming, Luke. Eric. Eric gonee kind of seems like Eric Anyway,

Luke Arrigoni 19:39
did you want to say the line over? I’ll pause? No,

george grombacher 19:41
no, we’re gonna leave that one in for sure. Let’s do it. Luke. Thank you so much for coming on. Where can people learn more about you and how can they engage with aircore

Luke Arrigoni 19:51
you know, you can reach out to me on my site, or you know, a lot of people just will directly message me on LinkedIn. You can message anyone that you find from from aircore on LinkedIn or through email, and, you know, all roads will lead back to me. So you could call us, of course, a 213 era core, but most people don’t have a phone or want to call anymore. So feel free to shoot us an email or a chat or just connect on LinkedIn.

george grombacher 20:14
Excellent. And I forgot to ask other certain size organizations that you prefer to work with.

Luke Arrigoni 20:21
Yeah, oftentimes, we’re aligned with people that are medium to enterprise size. If you’re a startup, that’s great. We often will offer guidance, you know, do a little bit of counsel that way, but typically, you’re you know, these these projects are six figure in nature. And so they they range in that medium to enterprise size companies. Got it, given the website again, please. That’s Eric, hoard calm ARR IC o r.com. Perfect.

george grombacher 20:49
Well, if you enjoyed this as much as I did show Luke your appreciation and share today’s show with a friend who also appreciates good ideas, go to air corps.com. It’s a R IC o r.com. And check out all the great resources and reach out to figure out what you can do with all that data. Thanks again, Luke. Thanks for having me. And until next time, keep fighting the good fight. We’re all in this together.

Transcribed by https://otter.ai