1 00:00:00,160 --> 00:00:03,060 Welcome to show 350 of data driven. 2 00:00:04,000 --> 00:00:07,460 In this episode, Frank and Andy interview Chris McDermott, 3 00:00:07,919 --> 00:00:11,475 VP of engineering at Wallaroo. Wallaroo helps 4 00:00:11,475 --> 00:00:15,014 customers operationalize machine learning to ROI in the cloud, 5 00:00:15,075 --> 00:00:18,910 in decentralized networks, and at the edge. It's 6 00:00:18,910 --> 00:00:22,670 a fun conversation on MLOps and the future of intelligence systems 7 00:00:22,670 --> 00:00:26,050 and model management. Now on to the show. 8 00:00:29,795 --> 00:00:33,575 Hello, and welcome to data driven, the podcast where we explore the emergent 9 00:00:33,635 --> 00:00:37,460 fields of artificial intelligence, Data science and of course, data 10 00:00:37,460 --> 00:00:41,300 engineering, the fundamental thing that kind of underpins it all. And with 11 00:00:41,300 --> 00:00:44,199 me on this epic road trip down the information superhighway 12 00:00:44,935 --> 00:00:48,455 Is my favorite data engineer of all time, Andy Leonard. How's it going 13 00:00:48,455 --> 00:00:52,055 Andy? Good Frank, how are you? I'm doing alright. I'm doing 14 00:00:52,055 --> 00:00:55,879 alright. We just, we're chatting 15 00:00:55,879 --> 00:00:59,660 in the, virtual green room about some of the logistical challenges we had, 16 00:01:01,800 --> 00:01:05,485 with Microsoft Bookings and how Kind of like you can only have, you 17 00:01:05,485 --> 00:01:09,325 know, like, remember that the pick any 2 triangle, right? Good, fast, and 18 00:01:09,325 --> 00:01:12,765 cheap? Yep. Yep. Like, we can only have 2 things, 2 features of what we 19 00:01:12,765 --> 00:01:16,110 needed to do. Right. Alright. 20 00:01:16,930 --> 00:01:20,630 Despite logistical challenges, we are excited here to have, 21 00:01:22,355 --> 00:01:25,655 Chris McDermott who is the VP of engineering at Wallaroo, 22 00:01:26,355 --> 00:01:29,815 and, he is a passionate, and intellectually 23 00:01:29,875 --> 00:01:33,700 curious professional With excellent communication skills, he 24 00:01:33,700 --> 00:01:37,540 loves hard problems, then he must have definitely loved the 25 00:01:37,540 --> 00:01:40,795 process to get on the show, And, 26 00:01:42,055 --> 00:01:45,655 have yet to meet 1 he couldn't solve somehow. Maybe we should get you, 27 00:01:45,655 --> 00:01:49,090 Chris, to help us with our scheduling stuff. Really? You visit 28 00:01:49,509 --> 00:01:53,229 later? Yeah. So welcome to the 29 00:01:53,229 --> 00:01:56,590 show, Chris. Thank you. Thank you. It's great to be on. It's nice to meet 30 00:01:56,590 --> 00:02:00,295 you both. Well, likewise. Likewise. So you're coming to us from the, 31 00:02:00,615 --> 00:02:04,375 Mile High City. That's right. Awesome place. It's, I was 32 00:02:04,375 --> 00:02:07,970 there once, for internal Microsoft 33 00:02:07,970 --> 00:02:11,569 conference actually. Oh, nice. And beautiful town, like, it was 34 00:02:11,569 --> 00:02:14,985 just really cool. I think it was the 2nd biggest 35 00:02:14,985 --> 00:02:18,345 event that in Denver history was the Microsoft thing. Wow. 36 00:02:18,985 --> 00:02:22,205 And they they literally ran out of hotel rooms like it was. 37 00:02:22,730 --> 00:02:26,490 Oh, wow. It was pretty wild. Yeah. I think it was, just 38 00:02:26,490 --> 00:02:30,170 before one of the big parties had a convention there. And, 39 00:02:30,650 --> 00:02:34,495 they Oh, yeah. Yeah. Yeah. Yeah. I was so I'm actually 40 00:02:34,495 --> 00:02:38,095 slated to head back there next year for a Red Hat 41 00:02:38,095 --> 00:02:41,930 conference, so we'll see Let's see if the hotel situation has 42 00:02:42,310 --> 00:02:46,150 improved. I think it's improved a little bit. The city's been growing a lot. So, 43 00:02:47,185 --> 00:02:51,025 Yeah. Lots of government. Isn't Denver the place that has, like, 44 00:02:51,025 --> 00:02:54,705 the large bear up against the conference center that 45 00:02:54,785 --> 00:02:58,189 Yeah. Yeah. Yeah. Yeah. Yeah. That's exactly right. A giant blue bear appearing in the 46 00:02:58,189 --> 00:03:01,549 window of the conference center. Yeah. I was there. And, 47 00:03:01,950 --> 00:03:05,314 and and I remembered that. That was the That was the first thing I 48 00:03:05,314 --> 00:03:08,295 remember. It was, I was there in 49 00:03:09,635 --> 00:03:13,230 2007 for a Kind of a Microsoft conference. It was 50 00:03:13,230 --> 00:03:16,830 a, Professional Association for SQL 51 00:03:16,830 --> 00:03:20,595 Server. That's what it was called back then. And, I was 52 00:03:20,595 --> 00:03:24,114 actually the first one I spoke at. I've spoken at a bunch since then, 53 00:03:24,114 --> 00:03:27,250 but 2007 in Denver was the first. And, 54 00:03:27,810 --> 00:03:31,350 yeah. Like, I echo what Frank said, beautiful city 55 00:03:31,810 --> 00:03:35,110 and, just very picture picturesque. Yeah. 56 00:03:35,755 --> 00:03:39,595 Yeah. The weather in the mountains are beautiful. Mhmm. Yeah. And 57 00:03:39,595 --> 00:03:43,035 it's funny, like, you know, on the East Coast, we talk about mountains, 58 00:03:43,035 --> 00:03:46,580 but It's nothing like that. Like Right. Yeah. We quite the 59 00:03:46,580 --> 00:03:50,340 same. You would laugh at what we call mountains. Yeah. Right. But 60 00:03:50,340 --> 00:03:54,114 I remember a Robin Williams bit Where he said something like that 61 00:03:54,114 --> 00:03:57,715 people he admired the people in Denver because they got to 62 00:03:57,715 --> 00:04:01,430 Denver and they looked at the mountains and went, Well, I can't say what 63 00:04:01,430 --> 00:04:04,890 he said, but he had a kind of an Elon 64 00:04:05,190 --> 00:04:05,690 moment. 65 00:04:08,894 --> 00:04:12,575 There's so many of those. There's so many No more. We're stopping right here. We're 66 00:04:12,575 --> 00:04:15,315 not going over those mountains. So, 67 00:04:17,839 --> 00:04:21,620 You're VP of engineering at Wallaroo. So tell us a little bit about Wallaroo. 68 00:04:22,639 --> 00:04:26,480 Mhmm. Plus you're also ex data robot too. That that's interesting. 69 00:04:26,480 --> 00:04:30,324 Yes. Yep. Exadata robot. Yeah. So I've been working in the machine 70 00:04:30,324 --> 00:04:33,845 learning and AI space for, about 7 years now, I guess, or 6 71 00:04:33,845 --> 00:04:37,620 years. And, it's been really fun. You know, it's, it's a 72 00:04:37,620 --> 00:04:41,080 good time to be in the business. There's a lot of development 73 00:04:41,139 --> 00:04:44,405 happening, very fast pace of change, which I appreciate. 74 00:04:45,105 --> 00:04:48,165 And, you know, Wallaroo has been really great. Like, 75 00:04:49,265 --> 00:04:52,145 the team is fantastic, and the people are wonderful. And it it's a lot of 76 00:04:52,145 --> 00:04:55,960 fun working, with people that you enjoy hanging out with and and you respect 77 00:04:55,960 --> 00:04:59,680 and everything, that's that's very important to me. That's awesome. But I also just 78 00:04:59,800 --> 00:05:03,275 I think the product is awesome. It's really, I think, 79 00:05:05,014 --> 00:05:08,294 playing well in the market. Like, we are focusing on making it as easy as 80 00:05:08,294 --> 00:05:11,540 possible to deploy And manage machine learning models. 81 00:05:12,000 --> 00:05:15,760 And the focus is really on any model using any framework and being 82 00:05:15,760 --> 00:05:18,900 able to deploy onto sort of any architecture, any hardware, 83 00:05:19,685 --> 00:05:23,465 and being able to leverage GPUs if you need them or different kinds of CPUs, 84 00:05:23,605 --> 00:05:27,285 different acceleration libraries that people have tailored to the different 85 00:05:27,285 --> 00:05:31,050 architectures. And, honestly, there are not a lot 86 00:05:31,050 --> 00:05:34,890 of other solutions that tackle those 2 problems for people. Right. 87 00:05:35,290 --> 00:05:38,935 A lot of the other companies that we're competing with, they are trying to be 88 00:05:38,935 --> 00:05:42,375 like an end to end solution or, like, really force people into, you know, their 89 00:05:42,375 --> 00:05:46,215 platform. So you train on their platform, you deploy on their platform, you manage on 90 00:05:46,215 --> 00:05:49,919 their platform. But it's very limiting in terms of what you can bring on to 91 00:05:49,919 --> 00:05:53,360 the platform and and being able to, deploy on the different types of 92 00:05:53,360 --> 00:05:57,135 architectures and, platforms and things like that. So it's really 93 00:05:57,135 --> 00:06:00,895 exciting. It's fun. I think that's really important that you bring up 94 00:06:00,895 --> 00:06:04,530 the CPU solutions. As I've been tinkering, 95 00:06:04,990 --> 00:06:08,670 you know, the past couple of years with, you know, with the 96 00:06:08,670 --> 00:06:12,205 different, different platforms that are out there, it's 97 00:06:12,604 --> 00:06:16,365 That's definitely a smaller market, but maybe it's emerging now. I'm 98 00:06:16,365 --> 00:06:20,145 just not sure. Mhmm. Yeah. I wonder yeah. Sorry. Go ahead. 99 00:06:20,289 --> 00:06:23,669 Well, I was gonna say, you know, a lot of the time people conflate training 100 00:06:23,729 --> 00:06:27,555 and inferring, which is, you know, sort of the 2 different stages. Like, 101 00:06:27,634 --> 00:06:30,595 1st, you have to train a model, but then you use the model to make 102 00:06:30,595 --> 00:06:34,115 inferences, which, you know, it's really like asking the model to make a prediction 103 00:06:34,115 --> 00:06:37,510 or you give it some input and it gives you some output. And, 104 00:06:38,070 --> 00:06:41,830 they're they're very, very different tasks. And just because, you know, 105 00:06:41,830 --> 00:06:45,555 like, you may wanna use some hardware GPUs for training Doesn't 106 00:06:45,555 --> 00:06:49,395 necessarily mean mean that you need the GPUs when you are in production and 107 00:06:49,395 --> 00:06:53,160 you're asking it for predictions. A lot of the time, you know, 108 00:06:53,240 --> 00:06:56,760 The model is small enough that you really don't need to, but there's 109 00:06:56,760 --> 00:07:00,520 so much hype. It it's hard sometimes to separate the hype from the, you 110 00:07:00,520 --> 00:07:04,264 know, The real stuff and Yeah. Yeah. The hype the hype 111 00:07:04,264 --> 00:07:07,784 machine is real. I mean, like, it's and and and I I wanna get your 112 00:07:07,784 --> 00:07:11,350 thoughts on, you know, I mean, I love generative AI. I'm not 113 00:07:11,350 --> 00:07:14,950 knocking generative AI, but it feels like it's taken all the oxygen out of the 114 00:07:14,950 --> 00:07:17,955 room for All the other kinds of AI. 115 00:07:18,495 --> 00:07:22,255 Yeah. Yeah. Yeah. Because there are a lot of, you 116 00:07:22,255 --> 00:07:26,030 know, great models. I like XGBoost is a very standard one. It's 117 00:07:26,030 --> 00:07:29,710 been around for, you know, a long time, meaning at least for, you know, 5 118 00:07:29,710 --> 00:07:33,470 or 10 years now or something. But, that really honestly solves so 119 00:07:33,470 --> 00:07:36,595 many problems, and it's such a Small, easy model to deploy. 120 00:07:38,095 --> 00:07:40,974 I I wish people would focus more on on that kind of thing rather than 121 00:07:40,974 --> 00:07:44,630 hype. Right. No. That's a good point. And I think you 122 00:07:44,630 --> 00:07:48,310 bring up an interesting point because not all not 123 00:07:48,310 --> 00:07:51,610 all AI workloads are created equal. Right? Obviously, there's, 124 00:07:52,455 --> 00:07:55,735 I heard this term the other day and I had to spit my coffee out 125 00:07:55,735 --> 00:07:59,095 because it was just so funny. Legacy AI. Yeah. 126 00:07:59,095 --> 00:08:02,680 Yeah. There's generative AI now. There's legacy AI. That's 127 00:08:02,680 --> 00:08:06,140 crazy talk. You know? And I was just like, 128 00:08:06,440 --> 00:08:09,615 wow. But, 129 00:08:10,235 --> 00:08:14,074 you know, because, you know, legacy AI, basically, 130 00:08:14,074 --> 00:08:16,895 you're not using deep learning, you're not using neural networks, 131 00:08:18,340 --> 00:08:21,480 Generally, you don't get a good boost from GPU's. 132 00:08:22,420 --> 00:08:26,040 Correct. Right. And that's something that when when you tell that to 133 00:08:26,515 --> 00:08:30,115 Even tactical decision makers, they they they 134 00:08:30,115 --> 00:08:33,554 kinda look at you like, you know, what sorcery is that? Like, you know, because 135 00:08:33,554 --> 00:08:37,190 they'll they'll They'll say, like, oh, we don't have enough GPUs. There's no budget for 136 00:08:37,190 --> 00:08:41,029 GPUs. Like, what what what types of workloads are you running? And I 137 00:08:41,029 --> 00:08:43,750 tell them, it's like, well, it's not really a concern for you. Like, you don't 138 00:08:43,750 --> 00:08:47,495 need them. Yeah. And you see, you know, 139 00:08:47,495 --> 00:08:51,015 the the the the people who are doing the actual data science, they're like, yeah, 140 00:08:51,015 --> 00:08:54,630 duh, that's what we're trying to tell you. Yeah. But you see, like, the leaders 141 00:08:54,690 --> 00:08:58,389 of these teams are like, like, you know, 142 00:08:58,690 --> 00:09:02,375 it's, now Just for my own education, 143 00:09:03,315 --> 00:09:06,834 there wasn't there something called RAPIDS, and it was an 144 00:09:06,834 --> 00:09:10,550 acronym that let you use GPU's For 145 00:09:10,550 --> 00:09:14,230 certain types of like XGBoost, I think was one of them. Random 146 00:09:14,230 --> 00:09:18,045 forest there. I don't know. Oh. You See, it's funny because 147 00:09:18,045 --> 00:09:21,884 it was an it's an NVIDIA thing and obviously it only optimizes on. But, 148 00:09:21,884 --> 00:09:25,380 like, it was I remember Hearing about 149 00:09:25,380 --> 00:09:28,899 it in 2019, and I'm thinking, wow, this is gonna change 150 00:09:28,899 --> 00:09:32,685 everything, and you haven't heard of it. Only, 151 00:09:32,685 --> 00:09:36,045 like, one ever per other person I met in the wild has ever heard of 152 00:09:36,045 --> 00:09:39,325 it, and he was at the same conference I was at where we heard about 153 00:09:39,325 --> 00:09:41,360 it. So I'm like, That's kind of unusual, 154 00:09:43,660 --> 00:09:47,259 but, you know, we gotta watch so 155 00:09:47,259 --> 00:09:50,935 fast, you know, and it's really hard to tell sometimes What 156 00:09:50,935 --> 00:09:54,615 what, which new developments are gonna end up being the future and which ones 157 00:09:54,615 --> 00:09:58,449 are gonna end up as dead ends? Right. You know, and even all 158 00:09:58,449 --> 00:10:02,209 the transformer stuff that that is powering GPT and and those similar types 159 00:10:02,209 --> 00:10:05,970 of models, I think that was originally written up in a 160 00:10:05,970 --> 00:10:09,635 white paper in, like, 2017 something. Mhmm. And it just kinda sat around for a 161 00:10:09,635 --> 00:10:13,395 while, and nobody paid a whole lot of attention to it until OpenAI really 162 00:10:13,395 --> 00:10:17,160 ran with it. So yeah. Pension is all you 163 00:10:17,160 --> 00:10:20,759 need. I think that's was that the paper? Sounds right. 164 00:10:20,759 --> 00:10:24,575 Yeah. And then we're gonna go. Oh, sorry. Go ahead. Sorry, 165 00:10:24,575 --> 00:10:27,935 Andy. I cut you off your point. No. I I don't wanna go too far 166 00:10:27,935 --> 00:10:31,375 downstream before I say cred boost for using the phrase I don't 167 00:10:31,375 --> 00:10:34,930 know. Oh, nice. Somebody with your 168 00:10:34,930 --> 00:10:38,769 credentials, you know, saying I don't know. That's that's super 169 00:10:38,769 --> 00:10:42,565 cool. So cred Honestly, there's way too much to know. There's no 170 00:10:42,565 --> 00:10:46,324 way anyone person could know that. I I like to joke. I 171 00:10:46,324 --> 00:10:50,070 haven't checked my phone or, like, news Feed in like half an hour 172 00:10:50,070 --> 00:10:52,630 and I'm like woefully behind now. Yeah. 173 00:10:55,029 --> 00:10:58,505 But it feels that way like in the whole Oh, no. It does. Yeah. Especially 174 00:10:58,725 --> 00:11:02,404 it was especially interesting when the whole drama on OpenAI, and I 175 00:11:02,404 --> 00:11:05,365 don't wanna go down that rabbit hole too far. But when all of that soap 176 00:11:05,365 --> 00:11:08,640 opera kinda unfolded Yeah. Yep. It was kind of like, 177 00:11:09,420 --> 00:11:12,940 what's the latest? Like, is he back? Is he gone? Is he working at Microsoft? 178 00:11:12,940 --> 00:11:16,320 Like, he did work at Microsoft for like 10 minutes, and now he doesn't. 179 00:11:16,460 --> 00:11:20,145 Like, Yeah. You know, at 180 00:11:20,205 --> 00:11:22,685 at some some point down the middle of it, it's like, call me when this 181 00:11:22,685 --> 00:11:26,450 is over, and I'll deal with the, things yeah. I'll 182 00:11:26,450 --> 00:11:30,130 check-in again. But that's just the human 183 00:11:30,130 --> 00:11:33,910 side of it, let alone the let alone technology side of it. 184 00:11:34,290 --> 00:11:38,095 So Operationalization. I think that's gonna 185 00:11:38,095 --> 00:11:41,855 be the buzzword. Obviously, chatty b t and JennyIs, taking 186 00:11:41,855 --> 00:11:45,209 all the air out of there. And I think the next buzzword It's gonna be 187 00:11:45,209 --> 00:11:48,810 operationalization. 1, because it's kinda 188 00:11:48,810 --> 00:11:51,790 hard to say, and I'm not gonna lie, I've had to practice. 189 00:11:53,264 --> 00:11:57,105 But, it's something that I think companies and organizations 190 00:11:57,105 --> 00:12:00,805 that adopt AI, whether it's legacy AI 191 00:12:01,480 --> 00:12:05,320 Or generative AI. They're gonna have to realize, like, it's one thing to build 192 00:12:05,320 --> 00:12:08,920 the model, and then it becomes a, okay, now 193 00:12:08,920 --> 00:12:12,495 what? Yeah. Yeah. Well and models 194 00:12:12,495 --> 00:12:16,255 really are just like any other software. It's not something that you just 195 00:12:16,255 --> 00:12:20,080 write once and you, you know, Throw it out there, and it runs forever 196 00:12:20,080 --> 00:12:23,920 without being touched. Right? All of it requires care and feeding, and 197 00:12:23,920 --> 00:12:27,705 and machine learning models are no different. So, I think 198 00:12:27,705 --> 00:12:31,145 part of it is, you know, how do you deploy it? And then, you know, 199 00:12:31,145 --> 00:12:34,825 how do you keep that that deployment up to date, you know, getting critical 200 00:12:34,825 --> 00:12:38,570 patches and vulnerability fixes and things like that. But also 201 00:12:38,570 --> 00:12:42,170 how do you monitor the model and how it's performing and how it's performing 202 00:12:42,170 --> 00:12:45,845 relative to the real world, Because the world doesn't stand 203 00:12:45,845 --> 00:12:49,305 still right. So even if the model was trained on some data and it was 204 00:12:49,524 --> 00:12:53,330 98% accurate when it was trained, as the world shifts and 205 00:12:53,330 --> 00:12:56,930 and the situation around it shifts, that accuracy will 206 00:12:56,930 --> 00:13:00,615 almost certainly start to degrade over time. So You need to monitor that. You need 207 00:13:00,615 --> 00:13:04,335 to know when to retrain the model. And you have to be kind of 208 00:13:04,335 --> 00:13:08,070 keeping track of, new training data. Right? So the 209 00:13:08,070 --> 00:13:11,350 the the new environment that the model is operating in, you need to be recording 210 00:13:11,350 --> 00:13:15,115 all of the the inputs and also paying attention to the ground truth of, You 211 00:13:15,115 --> 00:13:18,875 know, what was the outcome of that prediction that the model made? Was it accurate 212 00:13:18,875 --> 00:13:22,715 or not, after the fact? And and correlating that back into your 213 00:13:22,715 --> 00:13:26,380 training data So you can retrain the models and, you know, keep them going 214 00:13:26,380 --> 00:13:29,660 over time. And that's just, you know, assuming you're gonna be using the same model 215 00:13:29,660 --> 00:13:33,275 forever. But as we just finished talking about new models coming out all the 216 00:13:33,275 --> 00:13:36,795 time, new approaches, new techniques. So, yeah, it really is 217 00:13:36,795 --> 00:13:40,399 is something you've gotta pay attention to. It's an 218 00:13:40,459 --> 00:13:43,600 extremely Yeah. It's an extremely dynamic space. 219 00:13:44,060 --> 00:13:47,695 Mhmm. I've heard this called 220 00:13:47,695 --> 00:13:51,535 MLOps for the longest time. Mhmm. Mhmm. But I've also heard a new term 221 00:13:51,535 --> 00:13:54,964 kinda pop up on the radar called AI ops Mhmm. 222 00:13:55,214 --> 00:13:58,690 For this. What do you call it? I 223 00:13:58,690 --> 00:14:02,290 generally call it MLOps. You know, one, I 224 00:14:02,290 --> 00:14:06,115 I sort of per like, AI and ML, there's an 225 00:14:06,115 --> 00:14:09,635 interesting, you know, difference there in in terms of who uses the different terms and 226 00:14:09,635 --> 00:14:13,420 when they use them. For me, AI is 227 00:14:13,420 --> 00:14:17,100 more of a general term that I use conversationally. And most of the time if 228 00:14:17,100 --> 00:14:20,800 I'm trying to be fairly technical and specific, I'll usually revert to ML, 229 00:14:21,795 --> 00:14:25,634 Because in fact, most of these things are machine learning. AI is a much more 230 00:14:25,634 --> 00:14:29,394 nebulous concept, and I I don't even think everybody agrees on on what AI is 231 00:14:29,394 --> 00:14:32,780 or What the threshold would be, you know, if you're 232 00:14:32,780 --> 00:14:36,540 doing statistical analysis, I think most people probably would not call that 233 00:14:36,540 --> 00:14:39,680 AI. But there are a lot of machine learning models that do work that way. 234 00:14:40,205 --> 00:14:43,885 And and that's definitely, like, part of the gradient. You know? I've 235 00:14:43,885 --> 00:14:47,645 noticed that too. Like, there it is a gradient too. Like, there's not a, like, 236 00:14:47,645 --> 00:14:51,320 a hard, like, You know, typically it depends on the audience. Right? If they're 237 00:14:51,320 --> 00:14:55,160 if they're BDMs, business decision makers, they're gonna use 238 00:14:55,160 --> 00:14:58,835 AI. Yeah. They're technically focused people. They tend to prefer the term 239 00:14:59,135 --> 00:15:02,725 ML. Yeah. That's also been my experience Interesting. 240 00:15:02,975 --> 00:15:06,670 Quite often. So I like MLOps because, one, 241 00:15:06,670 --> 00:15:10,190 it sort of grounds you a little bit more in that technical perspective. Mhmm. 242 00:15:10,190 --> 00:15:13,815 And, and it's sort of a like, To me, I think I came up 243 00:15:13,815 --> 00:15:17,015 through DevOps a lot of my, you know, first half of my career was DevOps 244 00:15:17,015 --> 00:15:20,829 and infrastructure and things like that. And, I 245 00:15:20,829 --> 00:15:24,670 think part of the appeal of the term MLOps is it taps into a 246 00:15:24,670 --> 00:15:28,370 lot of the DevOps, associations. Right? And 247 00:15:28,535 --> 00:15:31,515 Right. The concepts and the themes of DevOps, which is really about, 248 00:15:32,535 --> 00:15:36,375 merging different skill sets and breaking down silos and getting different teams to 249 00:15:36,375 --> 00:15:39,250 communicate with each other and And to collaborate more, 250 00:15:40,029 --> 00:15:43,709 being more dynamic. Not just, you 251 00:15:43,709 --> 00:15:47,384 know, putting software out there and and letting it run forever, but Keeping it up 252 00:15:47,384 --> 00:15:50,824 to date, monitoring it, recording the logs, you know, all of that kind of 253 00:15:50,824 --> 00:15:54,345 stuff, and and getting into a flow of continuous 254 00:15:54,345 --> 00:15:57,860 deployment, you You know, continuous integration, continuous testing, continuous 255 00:15:57,920 --> 00:16:01,700 deployment. And I think on the ML side, that's also where 256 00:16:01,839 --> 00:16:05,675 MLOps really shines and and is bringing those themes 257 00:16:05,675 --> 00:16:09,435 to the party, rather than a data scientist training a 258 00:16:09,435 --> 00:16:13,019 model, deploying it, and, you know, Throwing it over the 259 00:16:13,019 --> 00:16:16,620 wall to to, like, an operations team or something. It's 260 00:16:16,620 --> 00:16:20,045 getting all these different teams and skill sets to work together. It's 261 00:16:20,045 --> 00:16:23,645 building a continuous, you know, pipeline, with 262 00:16:23,645 --> 00:16:27,485 monitoring and and feedback loops and so on. So that's that's why 263 00:16:27,485 --> 00:16:31,149 I like MLOps. No. I like that too. So in order to prevent 264 00:16:31,149 --> 00:16:34,829 any hate mail come in or or but actually comments, AI 265 00:16:34,829 --> 00:16:37,250 ops is also used, I've heard, in, 266 00:16:39,135 --> 00:16:42,735 the telcos and network operators tend to have a term 267 00:16:42,735 --> 00:16:46,175 called AI ops, where they use AI to help operate their network. So that is 268 00:16:46,800 --> 00:16:50,259 Got it. It's it's a it's a namespace collision, 269 00:16:50,560 --> 00:16:54,160 which I've free further which I prefer MLOps for to avoid the namespace 270 00:16:54,160 --> 00:16:57,515 collision, plus all the reasons you said. You 271 00:16:57,535 --> 00:17:01,375 know, what's interesting is and I came from a software engineering 272 00:17:01,375 --> 00:17:05,010 background and, you know, and I'll be honest, I was not 273 00:17:05,010 --> 00:17:08,770 initially a big, believer in in DevOps, but 274 00:17:08,770 --> 00:17:12,450 kind of as time went on, I became a convert. But I think 275 00:17:12,450 --> 00:17:16,195 that, you know, when you look at how AI models, ML 276 00:17:16,195 --> 00:17:19,575 models, whatever, how they get operationalized. 277 00:17:21,315 --> 00:17:24,810 You look at it And I I often I often can tell 278 00:17:24,950 --> 00:17:28,250 who the fans of the new Star Wars movies are by using this analogy, 279 00:17:28,630 --> 00:17:32,465 because I'll say it's The 2015 Star 280 00:17:32,465 --> 00:17:35,924 Wars movie and the 1977 movie, DevOps. 281 00:17:36,304 --> 00:17:40,000 DevOps being kinda like the original, episode 4 And then the 282 00:17:40,000 --> 00:17:43,760 new, the the first new one, right? It's the same 283 00:17:43,760 --> 00:17:47,120 plot. I mean, the characters have changed, some things are 284 00:17:47,120 --> 00:17:50,945 different, But very effectively, it's the same plot. And, you 285 00:17:50,945 --> 00:17:54,625 know, some people will laugh like you did, and some people I can 286 00:17:54,625 --> 00:17:57,820 see will, Their their faces turn red. But, 287 00:17:58,380 --> 00:18:01,660 but I mean it's like it's it's the same plat plot. The names, the places 288 00:18:01,660 --> 00:18:04,715 have some have changed. But you're right. I mean, I think and there's a lot 289 00:18:04,715 --> 00:18:08,095 of lessons we can learn Yeah. In the ML community 290 00:18:08,235 --> 00:18:12,040 from the DevOps world. Right? Because, You know prior 291 00:18:12,100 --> 00:18:15,940 to DevOps, you know, the developers and operations had a 292 00:18:15,940 --> 00:18:19,515 very antagonistic relationship for the most part. I'm sure there's always 293 00:18:19,515 --> 00:18:23,055 exceptions. You know, I was I was joking that they would only meet, 294 00:18:23,595 --> 00:18:26,715 they only have to interact 3 times a year, and one of those was the 295 00:18:26,715 --> 00:18:30,540 holiday Christmas party. You know what I mean? And 296 00:18:30,780 --> 00:18:34,460 Yeah. But if you wanna deploy something in a far more agile 297 00:18:34,460 --> 00:18:38,285 way where they have to, you know, you put it In some extreme cases, every 298 00:18:38,285 --> 00:18:42,125 few hours, some new bit of code gets gets pushed up. That's obviously on 299 00:18:42,125 --> 00:18:44,800 on one fore end of the spectrum. But for the most part, you know, a 300 00:18:44,800 --> 00:18:48,400 couple times a month is not unreasonable. You have to automate that. You have to 301 00:18:48,400 --> 00:18:51,600 have processes in place. Yep. And I I see a day, and if that day 302 00:18:51,600 --> 00:18:55,395 has not already come, I would be surprised, That AI is gonna be the 303 00:18:55,395 --> 00:18:58,515 same thing or ML. Right? You're you're gonna have to get but to your point, 304 00:18:58,515 --> 00:19:02,080 right, this is a continuous process, You know? Yep. Yeah. 305 00:19:02,140 --> 00:19:05,660 We can't get away with, you know, you have this isolated team of data scientists. 306 00:19:05,660 --> 00:19:09,360 They they they kinda go off to their little area 51 type labs 307 00:19:09,420 --> 00:19:12,585 in secret. Right. I then come back with some model, 308 00:19:13,365 --> 00:19:17,125 and and I'm guilty of this too. I've done this. Right? Where I'm like, I 309 00:19:17,125 --> 00:19:20,500 built the model. I'm done. I did the math. I did the hard 310 00:19:20,500 --> 00:19:24,179 part. How do you get the play it? Not my problem. Not my 311 00:19:24,179 --> 00:19:27,700 problem. And it's funny, like Yeah. You know, I caught 312 00:19:27,700 --> 00:19:31,155 myself. Right. I caught myself doing that as I, you know, you 313 00:19:31,155 --> 00:19:34,995 know, doing that. Like recently, I had to I had to do a demo 314 00:19:34,995 --> 00:19:38,730 and I had to work on a kind of a It's basically a predictive 315 00:19:38,730 --> 00:19:42,410 maintenance type thing, and I took all this data, had the model, and I 316 00:19:42,410 --> 00:19:45,775 just said, here's the here's the link to the model, Have at it, 317 00:19:45,835 --> 00:19:49,035 pal. Mhmm. And then as I sent that, I was like, you know, I should 318 00:19:49,035 --> 00:19:52,695 probably be more involved in getting this on a race for it. 319 00:19:52,695 --> 00:19:56,320 Right. Yeah. Yeah. Yeah. Yeah. No. I think that's a big part of it. 320 00:19:57,260 --> 00:20:01,040 Another big part of it though is, scale, you know. And I think scale 321 00:20:01,740 --> 00:20:05,355 scaling of compute and, how How people were using compute and how 322 00:20:05,355 --> 00:20:08,655 much compute was required was a big part of what drove DevOps. 323 00:20:09,195 --> 00:20:12,555 You know, if you were a sysadmin responsible for a 100 324 00:20:12,555 --> 00:20:16,220 servers, That's, you know, challenging, but it's feasible. Like, you can do 325 00:20:16,220 --> 00:20:18,380 that. You can keep them all up to date. You can keep them all in 326 00:20:18,380 --> 00:20:21,580 sync with each other. Make sure they they all have the same patch levels and 327 00:20:21,580 --> 00:20:25,365 and so on. But you scale that up to a 1,000 servers? 328 00:20:25,905 --> 00:20:29,025 That gets a lot trickier. You try to go to a 100,000 or, you know, 329 00:20:29,025 --> 00:20:32,850 if you're doing Internet scale things like Google or Facebook or somebody, We're talking 330 00:20:32,850 --> 00:20:36,450 millions, tens of millions. And Right. That level of scale 331 00:20:36,450 --> 00:20:39,970 requires you know, everything has to be automated. Everything has 332 00:20:39,970 --> 00:20:43,145 to Has to work that way and it has to be resilient and it has 333 00:20:43,145 --> 00:20:46,684 to, you know, have automatic fail over and stuff. You know, there's the, 334 00:20:47,385 --> 00:20:50,570 x k CD where they're, You know, they get to a certain point. They're just 335 00:20:50,570 --> 00:20:54,170 roping off entire data centers and being like, alright. We're throwing that one away and 336 00:20:54,170 --> 00:20:58,014 moving on to the next one. And for AI, I 337 00:20:58,014 --> 00:21:01,774 think a lot of the same stuff is happening. When, you know, 10 years ago 338 00:21:01,774 --> 00:21:05,600 or so when when people were just getting started on this journey And as an 339 00:21:05,600 --> 00:21:08,960 entity, as a business entity, if you're talking about 1 or 2 use 340 00:21:08,960 --> 00:21:12,640 cases, you know, you can have humans curate that stuff and hand 341 00:21:12,640 --> 00:21:16,385 craft it, hand roll it, hand deploy it, and hand manage it. But 342 00:21:16,385 --> 00:21:19,585 if you're a a big enterprise company and you you wanna have hundreds of use 343 00:21:19,585 --> 00:21:23,424 cases in production or thousands or tens of thousands, there's just no way. 344 00:21:23,424 --> 00:21:27,250 You have to automate it. No. That that that's a that's 345 00:21:27,250 --> 00:21:30,850 an excellent point. Like, one way I've heard to describe is that if you're 346 00:21:30,850 --> 00:21:34,524 baking a loaf of bread for your family and friends Or loads of 347 00:21:34,524 --> 00:21:37,445 bread. You can do it in your kitchen. Right? You don't have to do anything 348 00:21:37,445 --> 00:21:40,804 special, but if you're the Wonder Bread Corporation or 349 00:21:41,125 --> 00:21:44,530 Mhmm. And you wanna deliver at that scale, that's no longer an 350 00:21:44,530 --> 00:21:48,210 option. Mhmm. And I think that we're at that point where and 351 00:21:48,210 --> 00:21:51,605 correct me if I'm wrong, where I think AI and ML adoption or AI 352 00:21:51,605 --> 00:21:55,445 adoption is still new enough where there's enough of naivete out 353 00:21:55,445 --> 00:21:59,210 there of, oh, we don't need to scale to that degree. Like, we don't need 354 00:21:59,210 --> 00:22:02,730 the production line. I think I think that's ending. I think we're getting close to 355 00:22:02,730 --> 00:22:06,570 the the end of that era, but that's kind of been my yeah. I think 356 00:22:06,570 --> 00:22:10,015 so too. Yeah. Because they're they're more and more, ML 357 00:22:10,015 --> 00:22:13,795 tools in everybody's toolbox. Right? So you were talking about telcos 358 00:22:13,855 --> 00:22:17,590 routing network traffic using ML models. That's not 359 00:22:17,590 --> 00:22:21,350 gonna be 1 model. Right? Like, with latency and and 360 00:22:21,350 --> 00:22:25,195 everything else, you're gonna need, you know, Very small. Lots 361 00:22:25,195 --> 00:22:28,875 and lots of very small models deployed on every router, every top of rack 362 00:22:28,875 --> 00:22:32,635 switch, every, you know, whatever 5 gs cell phone tower, 363 00:22:32,635 --> 00:22:36,300 whatever you're talking about. There are a lot of cell phone towers. So you're 364 00:22:36,300 --> 00:22:39,980 not managing 1 model. You're managing a fleet of models, right, across 365 00:22:39,980 --> 00:22:43,725 different geos and all kinds of things. No. That's that's an 366 00:22:43,725 --> 00:22:47,405 excellent point. Sorry, Andy. That's okay. It does seem to scale like that, 367 00:22:47,405 --> 00:22:50,899 though. Right? It's almost it's almost tectonic. There's 368 00:22:51,200 --> 00:22:54,980 a whole new layer going down. You know? That's that's the new surface. 369 00:22:55,679 --> 00:22:58,745 I noticed on the website, I I popped over to wallaroo dot, 370 00:22:59,785 --> 00:23:00,285 aiwallar00.ai. 371 00:23:03,625 --> 00:23:07,225 And I noticed a familiar looking, blurb just 372 00:23:07,225 --> 00:23:10,809 below the top of page. And it's familiar to me because, I 373 00:23:10,809 --> 00:23:14,190 started off in business intelligence. I'm still working in BI. 374 00:23:14,650 --> 00:23:17,870 And there's a note, 90% of AI 375 00:23:18,090 --> 00:23:21,805 initiatives Failed to produce ROI. And I saw this 376 00:23:21,805 --> 00:23:25,245 in, you know, it's very similar number, 85% in, 377 00:23:25,725 --> 00:23:29,570 in BI back in the day. It's probably still true. So where 378 00:23:29,570 --> 00:23:33,170 does that number come from? Well, I think it reflects a lot of 379 00:23:33,170 --> 00:23:36,955 things. You know? Some of them we were just talking about and 380 00:23:36,955 --> 00:23:40,395 and where MLOps is coming from is is, a lot of the failure 381 00:23:40,395 --> 00:23:44,175 modes were teams not really working with each other. Right? 382 00:23:44,950 --> 00:23:48,710 Somebody decided we should be doing AI, so they hired the data scientist. 383 00:23:48,710 --> 00:23:52,250 And the data scientist works in the corner for a while and, 384 00:23:52,815 --> 00:23:55,775 You know, 1, they don't have access to all the data. They don't know what 385 00:23:55,775 --> 00:23:58,895 the data is, where to find it, how to access it, how to clean it, 386 00:23:59,215 --> 00:24:02,900 what it means to the business. There there are a whole set of challenges there. 387 00:24:03,679 --> 00:24:07,200 And then, you know, they may train some models and and get something, you 388 00:24:07,200 --> 00:24:10,835 know, to a point where they think that it's gonna solve a problem. But Then 389 00:24:10,835 --> 00:24:14,355 you've got to work with an IT organization to stand up infrastructure. You've got to 390 00:24:14,355 --> 00:24:18,195 work with somebody to package the model and build, you know, an API around 391 00:24:18,195 --> 00:24:21,940 it or a UI of some sort And figure out how to deploy 392 00:24:21,940 --> 00:24:25,540 it, train people on how to use it, and and actually, like, 393 00:24:25,540 --> 00:24:29,105 somehow integrate it into your business process So that it's 394 00:24:29,105 --> 00:24:32,625 it's driving business outcomes. And all of those are really tough 395 00:24:32,625 --> 00:24:36,065 challenges. And all of them require breaking down those 396 00:24:36,065 --> 00:24:39,730 silos and getting a bunch of different People within an organization to 397 00:24:39,730 --> 00:24:43,350 talk to each other and communicate and to work together to solve something. 398 00:24:44,930 --> 00:24:48,525 I don't think ML or AI is is a magic wand that you just 399 00:24:48,525 --> 00:24:52,365 wave and magically provide value to a business. You've got to really 400 00:24:52,365 --> 00:24:56,070 think about What is your business doing? And, you 401 00:24:56,070 --> 00:24:59,830 know, machine learning at at heart, it it's 402 00:24:59,830 --> 00:25:03,210 really just like a a more efficient way of 403 00:25:03,635 --> 00:25:06,855 Making decisions, you know, faster and more accurately, 404 00:25:07,554 --> 00:25:11,155 and with less human input. And so you've got to look for places where your 405 00:25:11,155 --> 00:25:14,529 business can either save a lot of money or make a lot of money by 406 00:25:14,529 --> 00:25:18,210 being able to answer a a simple question repeatedly very, 407 00:25:18,210 --> 00:25:21,695 very efficient. And that sounds easy, but in practice, 408 00:25:21,995 --> 00:25:25,675 defining a business problem is often one of the hardest parts. So now I'm 409 00:25:25,675 --> 00:25:29,080 seeing even more parallels. Uh-huh. Yeah. 410 00:25:29,460 --> 00:25:33,059 You know, that was the problem we were trying to solve, with 411 00:25:33,059 --> 00:25:36,740 business intelligence as well. So didn't mean to cut you off. Sorry about 412 00:25:36,740 --> 00:25:40,245 that. No worries. So I yeah. I think I agree with you. It it there 413 00:25:40,245 --> 00:25:43,125 are tons of parallels there. I think there are a lot of similar lessons to 414 00:25:43,125 --> 00:25:46,420 be learned, and I think we are applying them in this In this space in 415 00:25:46,420 --> 00:25:50,260 ways that we've applied them to other spaces in the past. I also 416 00:25:50,260 --> 00:25:53,220 think there are technical challenges. You know, part of it is the field is moving 417 00:25:53,220 --> 00:25:56,975 so fast. So there's just this constant stream 418 00:25:56,975 --> 00:26:00,735 of of new frameworks, new models, new techniques, and you 419 00:26:00,735 --> 00:26:03,470 have to kinda stay on top of that. You have to be careful with your 420 00:26:03,470 --> 00:26:06,770 tool selection, to make sure you're not, you know, 421 00:26:07,070 --> 00:26:10,625 going whole hog into some tool. That sounds 422 00:26:10,625 --> 00:26:13,825 great today, but it's just not flexible, and it's not gonna be able to support, 423 00:26:13,825 --> 00:26:16,885 like, all these new things that are coming out. Yeah. 424 00:26:17,860 --> 00:26:21,620 Or that company could have internal internal political 425 00:26:21,620 --> 00:26:25,460 strife, which was crazy talk. Right? Cast Absolutely. Right. Cast 426 00:26:25,460 --> 00:26:29,174 doubt on their future. Alright. That would never happen. That would never 427 00:26:29,174 --> 00:26:32,615 happen. Sorry. Yeah. You were talking about privacy, which I think is 428 00:26:32,615 --> 00:26:36,430 another key thing. Yeah. Data residency, data privacy, see data 429 00:26:36,430 --> 00:26:39,810 security. You know, all of those things matter tremendously. 430 00:26:40,910 --> 00:26:44,605 And for for a business trying to, get 431 00:26:44,605 --> 00:26:48,365 value out of AI and ML. You know, a lot of it, depends on 432 00:26:48,365 --> 00:26:52,159 having good data and, Cleaning it and curating it 433 00:26:52,159 --> 00:26:55,919 and getting it ready for things. But then it it forces the 434 00:26:55,919 --> 00:26:59,679 the organization to really kind of do an inventory. What do we have? What's useful? 435 00:26:59,679 --> 00:27:03,215 What's not useful? Well, how much do we store? How much do we not store? 436 00:27:03,754 --> 00:27:06,174 How do we comply with various regulatory 437 00:27:07,195 --> 00:27:10,810 environments? Right? GDPR is is the big one everybody, you know, 438 00:27:10,810 --> 00:27:14,570 loves to throw out there. It's it's big and it's complicated, but, you know, 439 00:27:14,570 --> 00:27:18,250 things like that matter a lot. And And there's 300 +1000000 people behind 440 00:27:18,250 --> 00:27:21,625 that. They're covered or whatever. I think that, you know, that that 441 00:27:21,625 --> 00:27:24,905 is not only do they have a big stick, but they have a big arm 442 00:27:24,905 --> 00:27:28,679 that they can wave that stick wet. Yes. You 443 00:27:28,679 --> 00:27:31,960 know, if if a small country with, like, you know, 50 people in it, and 444 00:27:31,960 --> 00:27:35,640 that could something like GDPR, people would just walk around it. But I think 445 00:27:35,640 --> 00:27:39,415 that, a block with I've heard different numbers, but 446 00:27:39,415 --> 00:27:42,955 it's for, you know, pushing 4 to 500,000,000 447 00:27:43,255 --> 00:27:46,929 people. That's a huge that's a big enough market nobody can really ignore. 448 00:27:47,149 --> 00:27:50,990 Yeah. What's interesting is on the LinkedIn page 449 00:27:50,990 --> 00:27:54,269 for Wallaroo I love the website, by the way. I checked that out too. Thank 450 00:27:54,269 --> 00:27:57,715 you. It talks about decentralized 451 00:27:57,935 --> 00:28:01,695 networks Mhmm. And at the edge. Yes. What how would 452 00:28:01,695 --> 00:28:05,470 you define decentralized network? Yeah. This is a big new push for us that we've 453 00:28:05,470 --> 00:28:09,230 been focused on for, I mean, we've been focused on it kinda for the 454 00:28:09,230 --> 00:28:13,065 last year, but it was a lot of, development on on the back end. And 455 00:28:13,065 --> 00:28:16,205 we just released kind of our 1st edge features and product, 456 00:28:16,905 --> 00:28:20,125 in October, so it's kind of a new thing for us. But, 457 00:28:22,390 --> 00:28:25,610 As you think about ML and edge or ML and AI, 458 00:28:26,390 --> 00:28:29,610 and the the fleets of models that we talked about and all these use cases 459 00:28:29,745 --> 00:28:33,425 And, you know, telcos and and five g cell phone towers and all of those 460 00:28:33,425 --> 00:28:37,105 types of things, intersecting with data and data 461 00:28:37,105 --> 00:28:40,710 residency and privacy and security, It it really seems to 462 00:28:40,710 --> 00:28:44,230 indicate to me and and to us at Wallaroo in general that the 463 00:28:44,230 --> 00:28:47,669 future is lots and lots of models being deployed in lots of 464 00:28:47,669 --> 00:28:51,075 locations. And I think that one 465 00:28:51,075 --> 00:28:54,595 big sort of industry wide theme that I'm seeing is if the 466 00:28:54,595 --> 00:28:58,170 last 20 years, let's say, was the story of Everybody 467 00:28:58,230 --> 00:29:01,850 picking up from their colos and moving to the cloud and centralizing 468 00:29:01,990 --> 00:29:05,690 all of their IT, I think that the next 20 years are gonna be 469 00:29:05,975 --> 00:29:09,575 Not like deconstructing the cloud. I think the clouds are here to stay and they're 470 00:29:09,575 --> 00:29:13,335 gonna continue to grow, right, year over year. But there will be more 471 00:29:13,335 --> 00:29:17,129 of a push out to more edge computing environments. Cell phones 472 00:29:17,129 --> 00:29:20,330 are getting more and more powerful. Cars are getting more and more powerful. Like, there's 473 00:29:20,330 --> 00:29:24,169 more computer stuff happening, all over the place, and the compute 474 00:29:24,169 --> 00:29:27,985 available, the memory and the storage available is all through the roof compared to 475 00:29:27,985 --> 00:29:31,745 what it was 20 years ago. And, I think we're 476 00:29:31,745 --> 00:29:35,510 gonna see more push for smaller, more specific machine learning models, And 477 00:29:35,510 --> 00:29:38,950 they're gonna be pushed out to all these edge locations so that they can run 478 00:29:38,950 --> 00:29:42,310 close to where the data is. So you're not schlepping this sensitive data all over 479 00:29:42,310 --> 00:29:45,825 the Internet and other people's networks. Yeah. 480 00:29:45,825 --> 00:29:49,345 But, you know, you're taking advantage of of compute resources that you 481 00:29:49,345 --> 00:29:52,485 have local to the data and making very fast decisions, 482 00:29:53,870 --> 00:29:57,630 you know, very efficiently. So I I have to jump in 483 00:29:57,630 --> 00:30:00,815 because, you you just made me feel really good. 484 00:30:01,695 --> 00:30:05,455 About a year ago, I built a large server here 485 00:30:05,455 --> 00:30:09,040 at home, which I hadn't done in a decade. Actually, my my 486 00:30:09,040 --> 00:30:12,640 20 year old son built it. But he and he helped me with, 487 00:30:13,280 --> 00:30:17,095 with picking out the new shiny fast parts, on it because I was 488 00:30:17,095 --> 00:30:20,475 so out of practice with this such confessing. 489 00:30:20,935 --> 00:30:24,559 But, and it's really cool to see, you know, all of his All of 490 00:30:24,559 --> 00:30:27,880 his skills. He does edge. We just picked up the 491 00:30:28,000 --> 00:30:31,679 Raspberry Pis are back in stock, finally. Yep. And I just picked up, 492 00:30:31,679 --> 00:30:35,285 like, 3 for $35, You know, the 1 gig force. 493 00:30:35,285 --> 00:30:39,065 Yep. Anyway, super excited about that. One of the things I built 494 00:30:39,205 --> 00:30:42,870 at the time I built a box About a year ago, you 495 00:30:42,870 --> 00:30:46,710 couldn't do a local GPT or anything close 496 00:30:46,710 --> 00:30:50,295 to that. And I said, Eventually, we're 497 00:30:50,295 --> 00:30:53,495 gonna be able to do this. I I made that guess, and it was a 498 00:30:53,495 --> 00:30:56,935 guess. Yeah. But about 6 months later, about 6 months 499 00:30:56,935 --> 00:30:59,840 ago, All of a sudden, I started seeing these 7,000,000,000 500 00:31:00,380 --> 00:31:04,220 token machines showing up and it started clicking. 501 00:31:04,220 --> 00:31:07,795 It was like, holy smokes, you can do this. I did make one stupid mistake 502 00:31:07,795 --> 00:31:11,095 and he didn't catch me on it. I bought a 12 gig GPU 503 00:31:11,155 --> 00:31:14,530 because that's super crazy huge From 10 years 504 00:31:14,990 --> 00:31:18,770 ago. And that wasn't super crazy huge at all. No. No. 505 00:31:19,390 --> 00:31:23,155 No. But it's interesting. Now they're back now. They can run on, You know, on 506 00:31:23,155 --> 00:31:26,675 the 12 gigs. And like you said, you mentioned the CPU models. So I just 507 00:31:26,675 --> 00:31:30,260 learned a ton as I've been going through this. And, That 508 00:31:30,380 --> 00:31:34,140 it's it's very encouraging to hear that. I had not heard anybody 509 00:31:34,140 --> 00:31:37,919 say edge and running small ML models on the edge. 510 00:31:37,980 --> 00:31:41,605 That's, I mean, that's what we've been trying to do here. And I I love 511 00:31:41,605 --> 00:31:44,905 the redundant you know, the idea of a redundant array of whatevers, 512 00:31:45,525 --> 00:31:49,260 you know, MLs. It's almost like a swarm of MLs. I've heard, 513 00:31:49,580 --> 00:31:53,340 yeah. Yeah. Yeah. That's true. Right? And, you know, I think there's a lot 514 00:31:53,340 --> 00:31:56,865 of interesting stuff happening on the battlefields in Ukraine right now drones. 515 00:31:57,125 --> 00:32:00,705 And Right. That Yeah. Was also a fascinating space and 516 00:32:00,925 --> 00:32:04,445 very much, I think, heading in the direction of lots of ML running at the 517 00:32:04,445 --> 00:32:08,260 edge. It's it's funny you mentioned that. So I live in a DC area, 518 00:32:08,480 --> 00:32:12,160 and, I was at a government tech 519 00:32:12,160 --> 00:32:15,845 symposium about 2, 3 weeks ago now. And 520 00:32:16,625 --> 00:32:20,385 they were talking about that that, you know, edge is gonna be much more important 521 00:32:20,385 --> 00:32:23,740 in the future of warfare. And he said presumably 522 00:32:23,960 --> 00:32:27,480 elsewhere too. Right? He was permanent primarily a government in defense. It was definitely a 523 00:32:27,480 --> 00:32:31,174 military industrial complex, type of type of event. But he was 524 00:32:31,174 --> 00:32:34,855 explaining like, you know, in the past, you know, 20 years, 525 00:32:34,855 --> 00:32:38,600 we've not dealt with adversaries. We've 526 00:32:38,600 --> 00:32:42,200 only dealt with adversaries in in battle space conditions 527 00:32:42,200 --> 00:32:45,500 where it was, you know, we controlled the airwaves. 528 00:32:46,375 --> 00:32:49,735 Mhmm. And he, I think he used an interesting term. We 529 00:32:49,735 --> 00:32:53,515 had airspace and electromagnetic electromagnetic 530 00:32:54,295 --> 00:32:57,960 dominance. I was also like, Wow. Yeah. That was yeah. Yeah. I was, like, oh, 531 00:32:57,960 --> 00:33:01,100 that's interesting. So, like, the whole idea of these disconnected 532 00:33:01,560 --> 00:33:05,085 decentralized networks, I mean, I think 533 00:33:05,085 --> 00:33:08,225 you're I think you're spot on. It's the future for 534 00:33:08,684 --> 00:33:11,825 geopolitical reasons, but also just for, you know, 535 00:33:12,210 --> 00:33:15,890 Privacy and just kind of flexibility reasons. Yeah. The 536 00:33:15,890 --> 00:33:17,670 question I have though is, like, 537 00:33:19,305 --> 00:33:23,145 Organizations can barely manage the infrastructure they have now and barely manage 538 00:33:23,145 --> 00:33:26,880 the software they have now. What are they gonna do when the software starts Not 539 00:33:26,880 --> 00:33:30,640 thinking for itself, but, like, this becomes another workload Yeah. On 540 00:33:30,640 --> 00:33:34,385 top of that. Like, what Well, for one thing, that's why Wallaroo It 541 00:33:34,385 --> 00:33:37,985 is focused where we are, and we're trying to build this platform to help people, 542 00:33:37,985 --> 00:33:41,265 you know, with this capability of being able to deploy models and manage a fleet 543 00:33:41,265 --> 00:33:45,029 of them at the edge. Because, yeah, there aren't a lot of good 544 00:33:45,029 --> 00:33:48,870 solutions for that today. Yeah. Interesting. I I think the 545 00:33:48,870 --> 00:33:52,625 general answer to your question is probably some combination of cloud and edge. 546 00:33:52,625 --> 00:33:55,825 You know, like, it does make sense to centralize a lot of things, and it 547 00:33:55,825 --> 00:33:59,530 makes the the maintenance easier and, more efficient. And 548 00:33:59,690 --> 00:34:03,050 You can get some economies of scale and, you know, all that kind of stuff. 549 00:34:03,050 --> 00:34:06,730 But, we are gonna have to get good at managing a bunch of, 550 00:34:07,130 --> 00:34:10,885 disparate types of things in desperate locations. I think all of 551 00:34:10,885 --> 00:34:12,425 us. Interesting. 552 00:34:14,645 --> 00:34:17,784 So this is the part of the show where we'll switch over to 553 00:34:19,699 --> 00:34:22,359 The premade questions, and for your convenience, 554 00:34:23,380 --> 00:34:26,280 I will, paste that in here. 555 00:34:28,844 --> 00:34:30,705 Hopefully, paste it. And there we go. 556 00:34:33,484 --> 00:34:36,880 So You had an interesting career looking at LinkedIn. You were at 557 00:34:36,880 --> 00:34:39,839 SendGrid. You were then you were at DataRobot, and you said you made a switch 558 00:34:39,839 --> 00:34:43,674 into the the data world, which begs the question, How did you 559 00:34:43,674 --> 00:34:47,514 find your way into data? Did data find you or did you 560 00:34:47,514 --> 00:34:51,230 find your way to data? I I 561 00:34:51,230 --> 00:34:54,989 guess that is a good question. I think that, it was probably a 562 00:34:54,989 --> 00:34:55,969 little bit of both. 563 00:34:59,045 --> 00:35:01,925 Finding my way to data, I think that the beginning of the story is probably 564 00:35:01,925 --> 00:35:05,305 at SendGrid. And I joined SendGrid as a DevOps engineer. 565 00:35:05,840 --> 00:35:09,440 And to be honest, I had not really heard of SendGrid at the time. I 566 00:35:09,440 --> 00:35:13,120 knew a little bit about it, but it, you know, I didn't really understand what 567 00:35:13,120 --> 00:35:16,734 it was, too much with the scale. SendGrid, by the way, is now owned by 568 00:35:16,734 --> 00:35:20,494 Twilio. But they have an API for sending email, and 569 00:35:20,494 --> 00:35:24,194 they make it just really easy to integrate with, websites and applications 570 00:35:24,415 --> 00:35:28,180 and and software so you don't have to worry about SMTP and, you know, 571 00:35:28,180 --> 00:35:31,860 DKIM signing and all the other, like, gnarly bits of of 572 00:35:31,860 --> 00:35:35,595 email. Turns out that Sengrid had a 573 00:35:35,595 --> 00:35:39,135 ton of data. They're handling billions of emails a day, 574 00:35:39,355 --> 00:35:43,080 and, you know, there's a lot of metadata there. The the actual data of the 575 00:35:43,080 --> 00:35:46,760 email and and so on, the recipients and who to send it to and all 576 00:35:46,760 --> 00:35:49,900 that stuff. And so working in that space, 577 00:35:50,725 --> 00:35:54,325 I was dealing with tons and tons and tons of data. I mean, we 578 00:35:54,325 --> 00:35:58,085 had, we were using mostly MySQL, and we had these 579 00:35:58,085 --> 00:36:01,619 massive massive clusters. I think we had, 580 00:36:01,619 --> 00:36:05,380 like, 30 or 40, you know, schemas under management. Each 581 00:36:05,380 --> 00:36:09,025 schema was a cluster of anywhere from, Like, 6 582 00:36:09,025 --> 00:36:12,705 to 40 plus servers, Wow. 583 00:36:12,865 --> 00:36:16,560 You know, with lots of compute and everything else. So that was probably my 584 00:36:16,560 --> 00:36:20,240 1st foray into, like, really thinking about data as a first class 585 00:36:20,240 --> 00:36:23,700 citizen. And, and even to the extent of, like, 586 00:36:23,975 --> 00:36:27,815 You know, building an architecture around the data. Right? So 587 00:36:27,815 --> 00:36:31,335 that you can optimize the flow of the data, and being able to store it 588 00:36:31,335 --> 00:36:34,830 and process it and transmit it fast enough to keep up with, with the 589 00:36:34,830 --> 00:36:38,430 flow. And so, yeah, from there, 590 00:36:39,070 --> 00:36:42,694 you know, had a lot of fun, learned a lot of things about, startups 591 00:36:42,915 --> 00:36:46,275 about industry, about, DevOps and and all kinds of 592 00:36:46,275 --> 00:36:49,714 things. Management as well and leadership because that's where I first, 593 00:36:50,355 --> 00:36:53,930 started managing teams And then moved to data robot and, 594 00:36:54,630 --> 00:36:58,470 into the ML space. And then it was a whole another learning journey 595 00:36:58,470 --> 00:37:01,694 about, you know, data, 596 00:37:02,795 --> 00:37:06,555 engineering, feature engineering, transformation tools. How do you 597 00:37:06,555 --> 00:37:10,260 curate your data? And how do you really, like, know what you 598 00:37:10,260 --> 00:37:13,960 have and inventory it and, make it available 599 00:37:14,099 --> 00:37:17,480 to people within the business so that they can get value out of it. 600 00:37:20,234 --> 00:37:24,015 Interesting. Very much. So our next question 601 00:37:24,155 --> 00:37:26,655 is what's your favorite part of your current gig? 602 00:37:29,180 --> 00:37:32,940 I think it's actually, I'm gonna cheat and I'm gonna say I have 2 favorite 603 00:37:32,940 --> 00:37:36,665 things. And I I kind of always have I I 604 00:37:36,825 --> 00:37:40,345 Figured out this formula a while back, in terms of what 605 00:37:40,345 --> 00:37:43,705 motivates me. And it's one part the people that I work with 606 00:37:43,705 --> 00:37:47,290 and another part, the problems that I had yet to solve. 607 00:37:47,430 --> 00:37:50,870 So I wanna work with smart people. I I really don't like being like, feeling 608 00:37:50,870 --> 00:37:54,474 like the smartest person in the room. I much prefer to surround myself 609 00:37:54,474 --> 00:37:58,155 with people that are smarter than me and I respect and I can learn 610 00:37:58,155 --> 00:38:01,615 from. But that also, you know, I enjoy. Right? 611 00:38:02,450 --> 00:38:05,250 We spend a lot of time at work, so it helps to to enjoy the 612 00:38:05,250 --> 00:38:09,010 people that you're working with. True. So that's a big part of it. And 613 00:38:09,010 --> 00:38:12,705 then, finding tough problems, hard challenges. You know, if I 614 00:38:12,705 --> 00:38:16,085 don't have hard challenges to keep me, to keep my mind 615 00:38:16,305 --> 00:38:19,900 engaged and occupied, I start to get bored and, that's no fun. I 616 00:38:19,900 --> 00:38:23,500 prefer to to always have something new to to to, you know, be chewing 617 00:38:23,500 --> 00:38:27,280 at. So, yeah, good people, smart people, 618 00:38:27,339 --> 00:38:31,165 and hard challenges. That is that is really awesome. I feel the 619 00:38:31,165 --> 00:38:35,005 same way about about both of those things. The, for me though, I 620 00:38:35,005 --> 00:38:38,570 I, Trying to find people that are smarter than me is 621 00:38:38,570 --> 00:38:42,330 really easy. So I I enjoy that part a 622 00:38:42,330 --> 00:38:44,830 lot. Like Frank. Frank is smarter than me. 623 00:38:47,085 --> 00:38:50,925 Well, thank you. So 624 00:38:50,925 --> 00:38:54,710 we have a couple of, complete this sentence, questions. 625 00:38:54,710 --> 00:38:57,850 The first one is, when I'm not working, I enjoy 626 00:38:58,150 --> 00:39:01,830 blank. When I'm not working, I enjoy 627 00:39:01,830 --> 00:39:05,674 reading. I Enjoy movies. I go biking sometimes. 628 00:39:05,734 --> 00:39:08,954 That's part enjoyment, part exercise. You know, it's good for me, but, 629 00:39:09,810 --> 00:39:13,330 There's a lot of good, road biking in particular around Denver and a lot of 630 00:39:13,330 --> 00:39:16,609 beautiful scenery. So you can, you know, just ride for a while and find yourself 631 00:39:16,609 --> 00:39:20,055 up in the mountains or something, which is great. Yeah. 632 00:39:20,055 --> 00:39:23,195 Traveling, cooking, all these things are good. 633 00:39:24,695 --> 00:39:28,109 Our next fill in the blank is I think the coolest thing about 634 00:39:28,109 --> 00:39:30,210 technology today is blank. 635 00:39:33,775 --> 00:39:37,135 I I don't think it's necessarily something about today, but I think the coolest thing 636 00:39:37,135 --> 00:39:40,755 about technology is how it builds on itself. I remember 637 00:39:41,200 --> 00:39:44,740 Years years ago, I was studying for the CCNA exam, and 638 00:39:45,520 --> 00:39:49,060 that was such a formative moment for me to 639 00:39:49,280 --> 00:39:52,945 suddenly understand How networks worked all the way 640 00:39:52,945 --> 00:39:56,645 from the physical, you know, sending 641 00:39:56,705 --> 00:40:00,485 electricity down a copper wire, and it can be on or it can be off. 642 00:40:00,690 --> 00:40:03,970 And that's it. Right? And you can do that really, really fast. Switch from on 643 00:40:03,970 --> 00:40:06,950 to off, on to off, on to off, all the way up to, 644 00:40:07,730 --> 00:40:11,045 web 2.0 and and Ajax and, you know, Asynchronous 645 00:40:11,265 --> 00:40:15,025 JavaScript stuff happening in Google Maps. Right? And I can just drag my map 646 00:40:15,025 --> 00:40:18,645 around. It's just mind blowing. And, honestly, 647 00:40:18,785 --> 00:40:22,450 like, that That journey from the zeros and the 648 00:40:22,450 --> 00:40:26,050 ones up to Google Maps, that was, you 649 00:40:26,050 --> 00:40:28,390 know, what, 50, 60 years of, 650 00:40:29,445 --> 00:40:33,125 technology building on itself of people solving very small simple 651 00:40:33,125 --> 00:40:36,885 problems, but you add up all those small simple solutions and you get 652 00:40:36,885 --> 00:40:40,250 something incredibly complex And absolutely mind blowing. 653 00:40:41,589 --> 00:40:45,430 Excellent. Very interesting. The last, the 3rd and 654 00:40:45,430 --> 00:40:49,234 final, Complete the sentence. I look forward to the 655 00:40:49,234 --> 00:40:52,535 day when I can use technology to do blank. 656 00:40:54,194 --> 00:40:57,860 I I would love, a Personal assistant, you know, like 657 00:40:57,860 --> 00:41:01,620 Jarvis from from Marvel Comics or something or, I don't know, 658 00:41:01,620 --> 00:41:04,980 from I I'm big into sci fi and and things like that when I read. 659 00:41:04,980 --> 00:41:08,655 So, there are plenty of examples, but some kind of a smart personal 660 00:41:08,655 --> 00:41:11,855 assistant that, you know, I can chat with and it keeps track of my calendar 661 00:41:11,855 --> 00:41:15,560 and reminds me of appointments and, you know, when to call 662 00:41:15,560 --> 00:41:19,400 my dad and whatever else, stuff like that. I just think that's 663 00:41:19,400 --> 00:41:23,135 so cool. And I don't you know, with Especially with all the new LLMs 664 00:41:23,195 --> 00:41:26,955 and and GPT stuff that's happening, I don't think we're super far from that. So 665 00:41:26,955 --> 00:41:30,710 it's kind of exciting to me. No. You're right. Like, I 666 00:41:30,950 --> 00:41:34,630 you know, if you watched, you know, when I was a kid, Star Trek next 667 00:41:34,630 --> 00:41:38,444 generation was on, And the way that they were able to interact with the 668 00:41:38,444 --> 00:41:42,204 computer just through their voice. Yep. And I mean, the 1st Star 669 00:41:42,204 --> 00:41:45,805 Trek show had that too, but, like, the way the conversations I thought were more 670 00:41:45,805 --> 00:41:49,380 richer and more kinda interactive. Mhmm. Mhmm. We 671 00:41:49,380 --> 00:41:53,059 have a lot of that now. Yeah. I think some of the fundamental pieces are 672 00:41:53,059 --> 00:41:56,755 in place now. Yeah. It'll probably take a little while to put 673 00:41:56,755 --> 00:42:00,295 them all together and make it work right. But yeah. Agreed. 674 00:42:01,130 --> 00:42:04,750 So our next one is, share something different about yourself. 675 00:42:05,210 --> 00:42:08,890 But we, always remind guests that we're trying to keep our clean 676 00:42:08,890 --> 00:42:11,855 rating. Yeah. On Itunes. So 677 00:42:14,494 --> 00:42:17,930 I don't know. I think one of the more interesting things about my 678 00:42:18,410 --> 00:42:22,030 Journey is that I don't have a background, like a a degree 679 00:42:22,089 --> 00:42:25,849 in anything technical. I went to college and I got 680 00:42:25,849 --> 00:42:29,654 my undergrad Studying Greek and Latin and classics. And 681 00:42:29,654 --> 00:42:33,095 so it was mostly history, archaeology, languages, and things like 682 00:42:33,095 --> 00:42:36,780 that. And Computers have always been a hobby of mine and and I 683 00:42:36,780 --> 00:42:40,460 definitely did some computer science stuff in high school. I took 1 or 2 classes 684 00:42:40,460 --> 00:42:44,105 in college, but I didn't really make my way into that 685 00:42:44,265 --> 00:42:47,005 Professionally until a few years after college. 686 00:42:48,665 --> 00:42:52,320 And, you know, honestly, I I don't think it's hurt me at 687 00:42:52,320 --> 00:42:55,760 all. And in many ways, I think it's helped me partly 688 00:42:55,760 --> 00:42:59,440 because, you know, it it helps a lot with management and leadership, just 689 00:42:59,440 --> 00:43:03,065 to To kind of have a broad background and and understand, you know, 690 00:43:03,065 --> 00:43:06,045 different people and perspectives and and where they might be coming from. 691 00:43:07,120 --> 00:43:10,960 And I'm sure that some of the languages, you know, studying languages helped me 692 00:43:10,960 --> 00:43:14,295 picking up computer languages as well. I think there are a lot of similarities in 693 00:43:14,535 --> 00:43:18,135 In, human languages and and computer, you know, programming languages. But 694 00:43:18,295 --> 00:43:22,135 What? Yeah. But, yeah, it is somewhat unique, and I don't run 695 00:43:22,135 --> 00:43:25,660 into too many other classics majors, At, you know, tech startups. 696 00:43:26,360 --> 00:43:30,040 I could definitely see the convergence, especially now when we're talking about 697 00:43:30,040 --> 00:43:33,855 LLMs and the like. Right. You know, the the 698 00:43:33,855 --> 00:43:37,535 nearest neighbor algorithms and all of that that are that are being applied 699 00:43:37,535 --> 00:43:41,120 because my understanding is that's that's, You know, that's how that 700 00:43:41,120 --> 00:43:44,960 works as it picks the next best word Right. You know, in a in a 701 00:43:44,960 --> 00:43:48,720 sentence. And so syntax and grammar and all of the things you 702 00:43:48,720 --> 00:43:52,235 studied in-depth, That should be very helpful. 703 00:43:52,615 --> 00:43:56,215 Yeah. No. That that's awesome. There 704 00:43:56,215 --> 00:43:57,995 is that good value in, 705 00:43:59,940 --> 00:44:03,700 like a classics education. I I went to Jesuit 706 00:44:03,700 --> 00:44:07,535 High School and Jesuit College, you know. Mhmm. I was forced into studying Latin 707 00:44:07,535 --> 00:44:11,214 and things like that, like, didn't do it voluntarily. I'm not gonna 708 00:44:11,214 --> 00:44:14,654 admit that, not do that. But but like as I get older, like, it's 709 00:44:14,654 --> 00:44:18,060 definitely like, Oh, I get this. Like, you 710 00:44:18,060 --> 00:44:21,740 know, especially when dealing with a lot of lawyers, there's a lot of Latin in 711 00:44:21,740 --> 00:44:25,445 that. And so I'll hear them, like, you know, Excuse some 712 00:44:25,445 --> 00:44:28,505 words. I'm like, I think I know what that means. Yeah. 713 00:44:31,205 --> 00:44:33,225 Audible sponsors data driven. 714 00:44:34,900 --> 00:44:38,660 And you mentioned you read a lot. Do you do audiobooks and 715 00:44:38,660 --> 00:44:41,800 sci fi? Do you have any recommendations? Yeah. 716 00:44:43,305 --> 00:44:47,005 There was a really good book that I read recently. Like, this is maybe 717 00:44:47,465 --> 00:44:50,925 a year ago or something, but, best book I've read recently. 718 00:44:51,145 --> 00:44:54,789 It's, The title of the book is called Seeing Like a State, 719 00:44:55,089 --> 00:44:58,549 and it's by, James c Scott. The the longer 720 00:44:58,609 --> 00:45:02,214 subtitle is, something like how Some 721 00:45:02,214 --> 00:45:06,055 schemes to improve the human condition have failed or something like that. But, 722 00:45:07,095 --> 00:45:10,635 it talks about this concept of legibility and how a lot of 723 00:45:10,900 --> 00:45:14,680 The developments over the course of the enlightenment, the industrial revolution, 724 00:45:14,980 --> 00:45:18,280 and, in the last few 100 years in in 725 00:45:18,685 --> 00:45:22,445 Our society have been primarily 726 00:45:22,445 --> 00:45:25,744 driven by the centralization of power in states 727 00:45:26,550 --> 00:45:30,390 And the state needing to administer all of these people, 728 00:45:30,390 --> 00:45:33,370 taxes, lands, land ownership, and all these different things. 729 00:45:33,955 --> 00:45:37,715 And, you know, as part of, like, the the enlightenment, the 730 00:45:37,715 --> 00:45:41,075 scientific revolution, we all got very enamored with, like, 731 00:45:41,075 --> 00:45:44,869 rational thought and Logic and and all of this stuff. And 732 00:45:44,869 --> 00:45:48,710 we thought, we're understanding the principles of the universe. We can predict 733 00:45:48,710 --> 00:45:52,305 the motions of the planets and all these things. Well, we can solve all these 734 00:45:52,305 --> 00:45:55,765 problems about, you know, around human civilization and humans as well. 735 00:45:56,625 --> 00:45:59,825 And in a lot of cases, it failed. Right? And we didn't know as much 736 00:45:59,825 --> 00:46:03,450 as we thought we did. And one of the sort of basic, 737 00:46:04,070 --> 00:46:07,910 like, premises of the book, I guess, or arguments that it's trying to make is 738 00:46:07,910 --> 00:46:11,734 that we routinely Underestimate, the 739 00:46:11,734 --> 00:46:15,275 complexity of the natural world and how necessary it is. 740 00:46:15,575 --> 00:46:19,240 And we think we can Simplify things and strip out all these 741 00:46:19,240 --> 00:46:22,940 variables and go, you know, monocultures in our in our agriculture, 742 00:46:23,000 --> 00:46:26,695 for example, and do industrial scale agriculture. You need 743 00:46:26,695 --> 00:46:30,455 timber for building ships. Great. We'll just plant Norwegian pines in straight 744 00:46:30,455 --> 00:46:34,220 rows. This is gonna be great. It's so predictable. We know exactly what, 745 00:46:34,300 --> 00:46:37,280 You know, an acre of that will yield after 10 years. 746 00:46:38,540 --> 00:46:41,500 But it turns out you can't strip out all the variables because the whole thing 747 00:46:41,500 --> 00:46:45,214 falls apart. You need the complexity of the ecosystem to keep all those trees 748 00:46:45,214 --> 00:46:48,974 healthy. And so all that predictability you thought you had 749 00:46:48,974 --> 00:46:52,620 disappears within a couple of generations because, it can't 750 00:46:52,620 --> 00:46:56,380 sustain itself. Wow. So, anyway, it it's a very, like, 751 00:46:56,380 --> 00:47:00,060 complicated book. I'm not really doing it justice, 752 00:47:00,060 --> 00:47:03,345 but I definitely recommend it. Interesting. It's on Audible. 753 00:47:05,325 --> 00:47:07,265 Yeah, yeah, so definitely check it out. 754 00:47:09,960 --> 00:47:12,460 The show. So if you go to the date is ribbon book.com, 755 00:47:13,800 --> 00:47:17,340 you'll be routed to an Audible page. And if you choose to get a subscription, 756 00:47:18,125 --> 00:47:21,665 to Audible. You will give us 757 00:47:21,805 --> 00:47:24,525 you'll get a free book, and then we'll get like a little bit of a 758 00:47:24,525 --> 00:47:28,190 bump on the head, and pat on the back, and Probably enough to 759 00:47:28,190 --> 00:47:31,970 buy a cup of coffee. It started Which will share. Which will 760 00:47:32,109 --> 00:47:35,250 share. Yes. Yes. And the final question, 761 00:47:36,494 --> 00:47:40,255 Where can people learn more about you and Wallaroo? And they even 762 00:47:40,255 --> 00:47:44,030 made that rhyme. Yeah. Great. I 763 00:47:44,030 --> 00:47:47,790 think the best place to go is the Wallaroo website, which, as 764 00:47:47,790 --> 00:47:50,850 Andy mentioned earlier is wallaroo.ai. So wallar00.ai. 765 00:47:54,335 --> 00:47:58,015 And we've got a ton of great stuff on there. Lots of, you know, 766 00:47:58,015 --> 00:48:01,695 documentation and and white papers and, tutorials and things about the 767 00:48:01,695 --> 00:48:04,829 product and what we're doing there. And for myself, 768 00:48:05,630 --> 00:48:09,250 I'm on LinkedIn. That's probably the easiest place to find me, Chris McDermott. 769 00:48:09,309 --> 00:48:12,849 And, I think I even have that as my, like, LinkedIn 770 00:48:14,005 --> 00:48:16,984 Profile name or whatever sits in the, you know, in the URL. 771 00:48:17,444 --> 00:48:21,285 Cool. It is, actually c s m 772 00:48:21,525 --> 00:48:25,309 c s McDermott. Okay. Well, thank you. Close. I was just looking at 773 00:48:25,309 --> 00:48:28,750 it, and I was also looking at the website. It's a very nice website. Thank 774 00:48:28,750 --> 00:48:32,235 you. Great design. And, although I can't design 775 00:48:32,695 --> 00:48:36,455 great websites, when I look at one, I can tell whether it's great or 776 00:48:36,455 --> 00:48:40,109 not. Me too. Me too. Same boat. I can't do it myself, but I definitely 777 00:48:40,109 --> 00:48:43,950 appreciate it. I I can't cook, but I appreciate a good meal. There 778 00:48:43,950 --> 00:48:47,535 we go. Yeah. That's it. And with that, we'll let 779 00:48:47,535 --> 00:48:51,215 Bailey finish the show. Thanks, Frank and 780 00:48:51,215 --> 00:48:54,895 Andy. And thank you, Chris, for putting up with our broken 781 00:48:54,895 --> 00:48:58,640 calendaring system. Satya should really look into that 782 00:48:58,640 --> 00:49:01,300 now that the drama around open a I is over. 783 00:49:02,179 --> 00:49:05,939 Well, over for now at least. Maybe g p t 784 00:49:05,939 --> 00:49:07,159 five can fix it.