happy-hour-nov-27-2020.mp3

[00:00:07] Oh, yeah, let's see what's going on. Here's my idea working. Oh, hi. Yeah, what's up, everybody? Welcome to the @TheArtistsOfDataScience. Data says happy hours. Hope everybody is doing. The waiting room is popping off. What's up? It's been an amazing week over here at the Artist. Data Signs release an awesome episode on Monday with Jeff Chrysler. Hopefully you guys didn't get dupa Black Friday deals. Speaking of which, it is November. Twenty seventh, Black Friday and super happy to have you guys here. Damn, the the it was popping off. This is awesome. How's everybody doing? All right. What's going on? We got so many people. Whalan, Giovana, Ashed, Naresh, man, so many people. I can't see what's going on here. All right. Cool. Welcome, everybody. Welcome back to another epic epic installment of the @TheArtistsOfDataScience AIs. Happy hour, Nicholas Lothal. Wow, this is amazing. Dave Languor is in the building as well. We got a buddy of mine who is it's like 3:00 a.m. in India and this guy is here making it happen, getting after a man. So I'll open the floor up to him first. But welcome, everybody. Super happy to have you guys here. A lot of new faces, man. So this is really awesome to see. Yeah, man. Mabern, how are you doing?

[00:01:29] Hi, Betty. Hi. Hi. How are you doing. Awesome. Happy to see you.

[00:01:36] Yeah. Super excited to have you here as well man. Yes. So it's three thirty a.m. for you. I know you probably want to get to sleep, so I'm happy that you're doing what you got to do to to get the help that you need. So I'm happy to to assist you, man. What can I help you with?

[00:01:53] Uh, yes.

[00:01:54] So, uh, I mean, let's get started and then I will ask and then start asking questions.

[00:02:02] Yeah, for sure. Man that is OK too. Christian what's up Nicholas. Dude, so good to see you, man.

[00:02:09] Yeah. Yes. Well it's another great great cast. Was like I wondered if it would be a little faster because of Thanksgiving, but I guess not everybody's here.

[00:02:18] Everybody's on lockdown. Nobody's Black Friday shopping. It's it's all just online. So people are just chillin at home. Man. Dave How's it going. Nishant Nearish I am here Machree chan Adam Jennifer shared injuries. Giovana of course is here. Whalen's in the House. So yeah man floor is open to anybody who wants to ask a question. Whoever wants to take the floor, go for it while that person is asking their question. If you just want to hold your place in line just type. I have a question right there into the chat and that will secure your place in line. So Wailin, how's it going, man? I know Weland was actually introduced to me by Maya Grossman, my assistant.

[00:03:00] Yeah, it was a I was I talked to her a little bit and she recommended me to talk to you and yeah, that's like happy hour thing. Like, it looks great.

[00:03:09] Yeah, definitely. So I'm happy to answer any questions. Have you got any? If not, then I want to see what's going on with Giovana and Monica and I thought Dave was here. David Languor. Yes he is. There he is. I'm sorry. There's just so many faces on my screen right now. We've got Cheerios here as well. Um, yeah. I'm super, super excited to have you guys here. Has your holidays been.

[00:03:30] Yeah. A lot of fun. Yeah, yeah, yeah. Yeah. I got go outside.

[00:03:35] It's like in there got NLP Thanksgiving dinners and all that stuff. Right man.

[00:03:40] Is that Ray that I see where you go. You're just on my screen there is great. Oh nice. Yeah. I've been following you for two weeks now. Tablo is kind of how I got started in, in the Data world. So I love to see the little tips and tricks you throw out there for Tablo.

[00:03:54] Appreciate your content and thanks man. So Ray, where he's running from and where are you located?

[00:04:00] I am in central Pennsylvania and the U.S. not too far from Hershey. Everyone knows Hershey chocolate.

[00:04:06] So about fifteen miles from there, I said, well, awesome, I'm glad you can join the super happy to have you here. I do see a question in the chat from Sheridan regarding getting started in Data science and how my podcast got started. Supinated everything. Will Ossman go ahead and meet yourself shared and you can just give us a little bit of background about yourself. Sheridan Oh yeah.

[00:04:28] Sorry about that guy. I'll get mine over here and try to get into my Uber and everything. Crazy day. I am currently and I'm currently in our that you Chicago and I'm just trying to supplement what I'm studying with and Data science and hopefully combine that into being an awesome career. So like I said, I'm really new and I'm trying to get to work and I just really appreciate being in the space. I just want to if there's any support or a welcoming offer, if there's any support or what, welcoming or. Like to observe, I suppose I'm going to get started and if any any kind of just general overview.

[00:05:10] Yeah, I mean, lots of tips and advice. That's a very, very big question to ask. But I'm just going to assume that since you're working on a Masters in Data Science, you've probably been exposed to all of the foundational technical knowledge, all the tools, mathematics, stats, programing and all that stuff. Right. So you've been exposed to all the theory you've put on the class and given all the tools, but you probably don't have much experience really applying that. Right. And that sounds about right. All right. So you need to kind of bridge that gap. Right? And there's always that that cycle of you need to get experience to get experience. Right. But you can create an experience. You can create your own apprenticeship in Data science if you've already learned the basics of the foundations. And the way you do that is by creating a really well thought out, well crafted project. And what I mean by well thought out, well crafted, is that it's got to be more than just like the datasets that are present on scalar and data says those toy data sets. Right. You've got to find a unique original project, not necessarily groundbreaking, earth shattering, but just something that's in line with your interest and in line with, you know, where you want to go with your career. So, for example, if you're trying to get into ecommerce, maybe you want to do a project that's a recommendation system, project return prediction project or something like that. So that's the next step. I would say. I'd love to hear what Dave Langer or Monica or Giovana has to contribute to that.

[00:06:38] I guess I'll go and go first. I think the probably the single most important thing is create a functioning definition of what Data science means to you in terms of what you want to do professionally. Unfortunately, if you're familiar with object oriented programing, Data science is now object. Everything is Data science, ranging from creating dashboards to keep learning computer vision for self-driving cars, all that kind of stuff. So if you want to create a project portfolio, which is a really good idea to things that are going to differentiate yourself, one, make sure that it's relevant to a business domain that you're interested in, just generally speaking, a general classifier. Titanic, which I love, by the way, everybody knows I love that. That alone is not going to be enough. If you want to work in supply chain analytics, study that build a project around that or marketing or whatever it might be, and also do research, go to these big companies where you might want to work, look at their job postings, cross-reference the kinds of skills and activities that they're listing, and that kind of gives you a more general sense of what the jobs actually are. So, for example, Facebook has a data scientist title, but if you look at the job descriptions, predominantly, they look a lot like a technical data analyst at another company. So just be aware of all that and plan appropriately because you just can't say I study data science and I'm a scientist. It doesn't really mean that much anymore because things are a little bit more targeted based on what kind of company, what kind of role you want to work for.

[00:08:04] But you that landscape, how do I find out what problems are solving in those industries?

[00:08:12] Yes, that's exactly what I was going to say to go off of what Dave was talking about, the domains of you find the domain that you're interested in, that supply chain and their financial industry, health industry, figure out what kind of problems those industries are trying to solve and focus on solving that problem. So while you're working on the project, no matter what cool tools are using or what technologies that you are working with, you just want to really hone in on what the problem that you're solving is and what that solution is.

[00:08:45] And Google is your friend. If you if you Google supply chain analytics, you're going to get back at MIT. Course ADEX that you can take, for example. That's a great way to actually learn about what the problem domain is and what kinds of tools and techniques they use. So Google it. Start with Google.

[00:09:03] You're very comfortable using Google about search and do searches for white papers of PDF Giovana. Yeah, go for it.

[00:09:09] Yeah, I, I love the things that I mentioned. They found Monica. I would like to suggest there is work here and be coherent when with all your projects and is the essence of armonica and they've, they have said another and another idea.

[00:09:30] It could be if you want to be part of a company, for example, if you have to mind that I want to work for this company. I love the philosophy of the mission of this company of but maybe not just one or maybe two or three. It's great to follow these companies on their website, only clean up with a to people post a response to the answer to the post that they. Company and all of these things helped because the first thing that's, you know, more about the culture, you know more about the project of a company that if they call you for an interview, you more and more you'll know more about them. And another important thing is that you are in contact with people who is interesting in that is interested in this company, too. And it's a way out to spotlight yourself with this company. So I think this is another way of knowing more about what is happening around that field. That is very important, because if you choose, for example, health care, you are there knowing more about this field. And at the same time, you knowing more about what Brilliants are doing. These companies, you can give some ideas in answering these posts, for example. This is something I could absolutely be.

[00:11:03] I think a lot of the times when people are breaking into the field, they spend a significant portion of the time just focusing on the acquisition of technical skills without really exploring outside of that and seeing what the trends are in the industry. So if we wanted to go figure out how to get up on trends in the industry, Joe or Ray or Nicholas, how would we do that?

[00:11:23] What are you guys just go to sources for that sort of cheating that I'm close to all types of customers and projects, usually kind of bleeding edge stuff, so I can have a pretty good idea of what's going on or what's coming right. But I would say to people like myself, too, if you want to get an idea of where things are going, at least like I say, I have a clear path, I think, versus a lot of people who discouraged. I was basically doing more data science, took a military data architecture Data engineering reasons why.

[00:11:58] But that said that the pace of change in Data is in terms of what's going on, new techniques, new technology is accelerating every day. So it's it's sort of weird advice. On one hand, you definitely the same type of stuff. Another other and there's a lot of noise. I would say 90 percent of the stuff out there is noise.

[00:12:16] And it's just important, I think, to stay on top of new advances and continue mastering the basics. So what were the basics? The Blinder's courses, I think, provide a very good foundation, analytics, things like SQL, things like, you know, if you're going to get engineering, things like shell scripting and then just things aren't going to go away, get good at that. I think what you're going to do, the basics you can circle back and I think have a better actually a better sense of how new technological advances are going to be, whether they're going to be important or not.

[00:12:48] So I would love to hear from from either Nicholas Ray. What do you guys think?

[00:12:52] I think there's an interesting parallel and everybody kind of feels like there's a lot they're rock stars in here. So that's that's good. And something that Joe touched upon about keeping on keeping on top of advancements in technology. I think there's an interesting parallel with we talk about the business context and the business application, and that's actually keeping on top of business challenges. And I spent quite a bit of my time in there being the schmoozer and the salesperson and approaching people essentially trying to get by and trying to get money for some of our big projects and then handing it over to people who are much more skilled in the science part than I am. And I found some of the quickest way to get on with people is to learn to speak the language of the people that you want to get inside. So if you're interested in Data science from the point of view of games and marketing and you will all you want to break into Supply-Chain Analytics or anything, then speak the language of the decision makers in the area is so important because you want to be able to speak not in terms of not just in terms of this great model. And it's X, Y and Z, but speaking. What things are they interested in? Customer revenue? I think it sounds simple, but it's a great way to build trust and keeping on top of challenges that the organizations that you work with, it is is pretty important.

[00:14:14] They're building a repository of that key terms and vocabulary. So you got these ideas of what it is that they're talking about, the industry.

[00:14:22] Sorry, I cut somebody off and I'm going to add to, I think some of the more interesting people that I've hired, some of the more successful ones have been those that are kind of zigging or the people are zagging. Right. So I would say it's, if you can, one of the best hires ever made as a 19 year old college dropout who sent me a resume and he shows up in my office of the combat boots nose ring looking. I don't know if he is looking to rob me or looking to get a job or something, but in either case, you know, it's a bit of an unusual approach this year.

[00:14:52] Let's do a quick interview. And he blew me away. Is this technical competence, his ability to communicate? I think we're better than a lot of the senior engineers they've hired, at which point I was like, well. You have a job, when can you start and the thing that really caught my attention was his passion around engineering and photography, right. Like he was doing that in his spare time, I think, like the age of 12 or crazy. But, you know, he she demonstrated a passion, something he was interested in and was able to see maybe not terms that I was familiar with, but I could tell, like, this is something that he was really passionate about and took a lot of interest in. And I'll see if you can find projects that are interesting to you and really dove into those and publicize those. It really goes a long way. Don't do like people. There's definitely a bias against you, like the Can Sanford project. You can take Hanneke or whatever, and it's not a million times or you can find something interesting to you. You're also going to learn all the latest techniques and everything you need to learn to solve that problem just because you want to solve it. I think that might be one of the better ways to approach, if you're starting from scratch, that we should have invested interest in something and you're not being force fed a bunch of information that the technology for trying and technology, I think, is that if you are applying it, you're going to forget about it.

[00:16:07] So one more thing. One more thing. I would add just to build on top of my good buddy Joe, here is the technical virtuosity aspect of your project is just basically table stakes these days. Everyone's got a deep neural network. Everyone's got a history, whatever. Where you can really shine is in your communication. Make sure that your code is well read and well commented and your variable names are nice. And make sure that you have great crafted and finely crafted verbiage around your project on your GitHub pages. Maybe create a YouTube video with your face in it like this and recording it and use your verbal communication skills. Wrap that all up because hiring managers, they look at it is going to say somebody who knows how to do deep neural networks these days is kind of a dime a dozen. But somebody they can do that and communicate well.

[00:16:55] Speaking of skills that never go away, communication, empathy, problem solving, problem finding and critical thinking. Oh, my God. Greg Cucu in the house. And I get to see you again, Greg. Good to see you again, my friend. Happy to have you here because.

[00:17:12] Hey, man, what's up?

[00:17:13] Good, good. Good to see you, man. Let's let's move on to Melanie. Uh, Cannibalizes said for some very excited people, I'm excited to have you money. I don't know where you are my screen, but. Oh, there you are. How are you doing? Hi, how are you? I'm good, thank you. Well, you've got a ton of people here who can help you with any advice you need related Data saying so go ahead and ask away.

[00:17:35] I sure I it's my first time and that I was communicating with you through the LinkedIn, which I really appreciated. You mention it's going to be in a central time. So I was able to sneak away from my position, from my job. So yeah, I'm Alanya. I am a buyer from a kitchen or maybe patriotic as I understand myself, trained myself and Nikitin currently. But I'm interested in moving away because I feel now I'm ready to move and then join industry. So my main understanding is there's a lot of opportunities in the street. But just because I'm coming from academia, I'm not sure the best way that I can represent that. I am ready to get into the position. I'm here to learn from your expertize and knowing how you transition away from academia because I saw find myself. So that means I do not have that many resources in academia to help me get out. But yeah, and I would love to learn from anyone and everyone and the I don't that thank you for, for coming and thank you for asking questions.

[00:18:45] So do you think your biostatistician right now, is that what you said about affirmative action? Yes. OK, yes. I used to be a biostatistician once upon a time, so it's a very similar transition for me. I mean, it wasn't academia, but it was very research oriented, research focused role. And it's a very academic in that sense. And I think the biggest I think the biggest thing for me that the the gap that I had to bridge was understanding how business people think and as well as understanding how to write decent software. So that's like the two big C, I'm guessing you probably use a whole lot of SAS right now or a whole lot of. Ah, yes. Ah, yeah. So in terms of translating your experience with ah from academics to industry, I would probably toss it over to Dave, but Tom as well. I did not see Tom there. Tom is here. Tom's also a academic as well as Dr Tom in the house. So I'd love to hear what advice either one of you guys have for her.

[00:19:47] Oh shoot. I really don't know much about Salsinha. I always defer to my buddy Dave on that too, or Danimal. But if you can use it in production, it's a great tool.

[00:20:00] Has a commute. I think he's done, yeah. Go for it. So let me try to rephrase the question.

[00:20:07] Some lady said she has some experience in R, but in an academic setting and now is trying to bridge the gap from academia to the business world where Data science is primarily played. So what challenges do you see somebody facing coming from an academic background that primarily uses are in that research setting to somebody who's going to be using it in a business type of role?

[00:20:30] Just so we're clear, I've never been in academia, only in the commercial realm. So keep that in the back of your mind. I would think based on my experience, probably the single biggest thing to decide once again, and I know I sound like a broken record, is where do you want to work? What kind of role do you want? As Tom mentioned, one of my favorite buzz words that just gets me just going, which is production. If you if you want to write and build and maintain production, machine learning pipelines, generally speaking, you're going to need to get off and get on Python. That's just the expectation in the job market. Rightly or wrongly, that's just the way it is. So that's the first thing you have to decide is do you want to stick with our do you want to find roles that in the market are conducive to in our background, which there are many, many, by the way, don't believe the total python hype. There are plenty of jobs out there where they'll say, oh, you know how to do this in R, OK, no problem. But worth. But as I said before, if you want to do production computer vision with automated cars, you're going to need to learn Python. That's just the way it needs to be where it goes. So if you want to move into academia, that's to move out of academia in the industry.

[00:21:41] That's the first thing you need to decide. Do research, look at the companies, look at the jobs. For example, I've looked at 70 something job descriptions from Amazon and Facebook and they're all basically, look, if you got our python, it doesn't really matter. Can you actually do the work in most of the most of the skills are relatively I don't want use the word rudimentary because that's so derogatory, but relatively simple in a good way. So do you know a Data this tool, if you don't have a Data vista, will pick up Tablo or Pattabhi, probably Tablo, because that's the dominant tool, but it doesn't have to be powered by works just as well. Or how to visualize Data match that with your basic statistics or even more advanced statistics and then R and then say, OK, this is my bundled skillset, these are the things I can do and then go map that to jobs in various companies and then you can say, where's my fit gap? And generally speaking, you might be surprised that for example and Greg could probably attest to this as an Amazon, for example, if you want to get a data analyst position at Amazon, if you have ah, you have got data visualization, if you got statistics and if you can speak reasonably well the language of the business, you're probably going to be a pretty powerful candidate.

[00:22:49] Speaking of getting a job at Amazon, LinkedIn top twenty twenty cockier is actually an employee of Amazon. So what does someone need to do to get a job at Amazon?

[00:23:00] That's a cool question. I think Amazon should be put in the basket of any other company looking for foreign talent. One thing I'd like to say is I like to tell people is, you know, let's start demystifying companies like Google, Amazon, and in understanding that we are the talents that are driving their success. So there's really nothing special. If you think about any company that produces a product or service, they will come up with different problems to solve and depends on what your skills are. You will find an opening a job position that best fit your skills. I wanted to say something to Milana about academia.

[00:23:55] One thing I wanted to say is I think before you meet that job and they've made a very great point. I think if you start from the from the end, meaning where you want to be, what kind of problems you want to solve from a business problem statement, then work backwards to figure out what are the tools and skill set you need to use to solve these business problems. So in academia, you are deep diving into the theoretical realm, which is a very strong point. And if you can start building those bridges between all of this theoretical knowledge with real life problems these companies are going through, if you zone in on the ones that are of high interest, then you can work backwards to figure out what are the tools, including the ones that they were talking about. So when it comes to Amazon finding a job there, what are your skills or trying to do some research about what what are the problems this company is going through? And then these companies also they're also good at giving you how to. Prepare for the interview. So they have the fourteen leadership principles for Amazon. They have the start method. I use the star method to redo my resume. I prepared to fourteen leadership principles. I also trained, did some mock interviews, recorded myself, listen to myself, respond to the questions and the star method and then readjust it. That's pretty much it. There's no mystery to it.

[00:25:30] And also I say that. Yeah. And I also would say that I don't think that the experience that you've had as a biostatistician bioinformatics man is like gone to waste. Like you've got a skill set that is extremely valuable in industry as well. I mean, I didn't know how valuable writing hundreds and hundreds of statistical analysis plans were going to be until I realized that people in the industry don't document shit that well and have to get stuff in order. And so that that became a very handy skill set to have. So let's keep it moving along. If you have any other questions, just feel free to let us know. Somebody here is asking if they if it's necessary to get a Masters to have a good job as a data scientist. Looks like there's a ton of discussion about that. I'm willing that the answer is no. Is there a consensus among that, Monica? What do you think? Giovana, what do you think? Do you need? I mean, I have a master's degree in math and statistics, so I might be biased, but you don't want was the question of masters in Data science.

[00:26:35] Yeah. Who didn't exist when I was in college. I'm sure that didn't exist for a lot of others. Yeah. So that's something very new. But if you do have to your point, masters in mathematics, statistics, maybe computer science, something quantitative that's transferable, I think that would be beneficial as well, or even in another area that's not technically math related. But you can have that communication skills and writing skills that are transferable as well and and kind of learn the math.

[00:27:09] I think it would be more beneficial to have at least something math related, though, if the question was more like, is it necessary to do a master's to have a good job as a data scientist, so could be any Vasteras. But then, as we talked about earlier, they pointed out correctly that data science is sort of a loaded term at this point. So it's you're dealing with like a master's degree in an overloaded field. And so I would say just pick something you're interested in doing. I don't know. I mean, yeah. I mean, if you want to go work at Open Eye, for example, you probably need a PhD. Right, because you're a researcher. Yeah, I got here last week about this and it's interesting, all the different perspectives on Data science.

[00:27:50] And I write like you get the people who are working on GBG three, that's not the same thing, are going to be working out like an insurance company. Yeah.

[00:27:57] So there's there's like really Data sounds like an entire ecosystem made up of a bunch of different roles. Right. You've got your traditional data scientist, Rogich, data analyst, really got your data engineering. That's machine learning engineering. Then you've got like a research scientist type role, any role that involves heavy, heavy research. I'm willing to bet that you will need to have graduate training, graduate education, that all these other roles are really practitioners. You're a meteorologist. You're applying a certain tool to a certain problem. Do you need a Masters to do that? Know? Do you need to know your shit? Absolutely. You need to know how to do the job in order to have a good career because otherwise you'll end up getting fired. So you need you need a solid set of skills that you can demonstrate and just perform well in your job and then you have a good job.

[00:28:47] But I wouldn't say so. Yeah. Yeah, they've correct me if I'm wrong. I don't know if you've research that before. So I wasn't a research scientist. Typically do fish for a level person. This is because there are focusing on things that haven't been solved before. So they are heavily trained on statistical analysis and knowing how to test and retest and readjust.

[00:29:14] And then you have the applied scientists that are either PhD level or a master's level because they do have a very sound, you know, education on statistics. So I don't know if you've ever researched that days, but that's what I'm seeing on on the Amazon side.

[00:29:33] Yeah, that's that's pretty common in all the FANG plus companies that I've looked at, generally speaking of scientists, is in the title PhD is usually required. Sometimes you might see like masters with like ten, fifteen years of experience. But generally speaking, it's a PhD once again, it really boils down to like what what gets you excited? What do you want to do every day that really allows you to choose your own adventure in terms of your investments? Me personally, I used to like geeking out and I was like, slide the pizza under the door in the middle of the night. Well, I'm cut and go. When I was a software engineer, I used to. Me these days, when I really like doing is I like looking at Data and then trying to influence the direction of the company or the organization based on the insights. And to be honest, you don't really need that much advanced stuff to do that really don't need don't be the hardest. Hardest thing is the communication skills.

[00:30:21] Usually if you don't want to like you don't want to live a life where you're just trying to meet spec on a sheet of paper. So don't worry about meeting specifications as required in a job description. Just do the work, get the experienced hands on whether it's a personal project, whether it is working on the job, and focus more on developing the skills, developing the right mindset to think through a problem that's critical thinking, that is realizing that you don't need to employ the scientific method in some manifestation of itself to solve business problems. So focus on acquiring tangible skills rather than meeting specs on a totally great atlas.

[00:31:07] Let's talk about like the cruel fate of irony of the universe. I don't even have a masters degree. I teach graduate school. Yeah, I mean, I teach analytics data scientist. Right? So it's like and this is something I tell my students a lot is is exactly their point. It's like you can beat all the all the checklists. I can make an arbitrary checklist for a data scientist and you can try to make that. That's not going to make you a good data scientist, though. It's going to be like checking whatever. I just sat in front of you to pick something interesting, right? Yeah. You want to find work, if you like, and that you love. I mean, you've got a long career ahead of you. And there's nothing worse than seeing people who have been stuck in careers. They just really hate you.

[00:31:42] And I think the two prongs to that question as well, because as we've touched upon already, if the question is do I need a masters to signal to recruiters that I'm equipped technically to tackle the kinds of problems that I'll be engaged with, what we've already alluded to is probably not, because if you can build a portfolio of project projects, you're demonstrating that you have the ability. So it's not a little piece of paper that says you have massive science, but you have a portfolio that shows your second problem to a solution. You've had some kind of impact. So if you're looking at it from the do I need a master's to signal that I have experience to recruiters? And the answer is probably not, apart from some of the more specialized roles he's spoken about. So then the question then becomes, do I need a masters to land the technical skills that I would need to engage with the job? And the answer to that is in science, probably not, because there are a wealth of brilliant resources out there that you can tap into, whether that's mathematics, statistics, programing like it your fingertips. So I think when you unpack it in that way, I'm in the camp of you don't need a Masters, you don't need a degree in that thing that you want to learn to be good at it.

[00:32:52] Right. I like to hear Ray's perspective on this as well, because I haven't heard much from him. Bay Area can chime in, definitely go for it.

[00:32:59] And then so just to be clear, I am not a data scientist, so I'm here as a listener, but I have a lot of experience in databases and just databases and stuff in general. So basically I'm trying to live out exactly what you guys are talking about. So the reason I'm trying to learn this stuff is I want to solve more problems. I want to be able to have that influence that Dave is talking about and other people. So I'm not at this point pursuing a master's because I have the Data is I have the Data wrangling and I have all of these on a lot of the major elements that go into it, but not the not the fluency and or the algorithms. So I'm trying to build that out. And I know since that's already my career. Right, I'll be given an opportunity or I can even do it on the side. But I have access to good data to do it with and things like that within my company so I can try to build out there. And that's what I'm hoping to do.

[00:33:58] I love it. I absolutely love it. Let's see what other questions you got in here. So, uh, Krishna, there's a question about resources that I used to get up to speed on the business side when I switched over. Just read a lot of books. So the biggest one for me was Lean Analytics by Alister Kroll and Benjamin Yoskowitz. By the way, you can listen to my interview with Alison call on my podcast. There's another book called Data Science for Business.

[00:34:26] Those are like the two biggest ones. And then just reading like you go to any major companies blog or or any major player in that particular industry, and you go to the blog and just research like Data or analytics or machine learning, you'll get introduced to a whole host of like blog posts talking about the type of stuff that they're working on and that just serves to build a vocabulary. You can understand, OK, this is how they're using this thing in this way. So that's how I would do that. I just wanna take a minute real quick shout out.

[00:34:56] Are you nice to see you, man. I know it's super late for you in India or early, depending on I don't know what it was like. There's also Jacqueline and Jacqueline, Jacqueline, if you got a question, please go for it.

[00:35:08] Yeah, so thank you. Just first, I want to say Harp Harp, thank you for organizing this. So I'm trying to start in the field and I'm sure you'll be super helpful for me. So I big like Melania. I just finished a Masters in Mathematics and I'm trying to bridge to the least bridge between my theoretical knowledge and I come more in a business setting. So maybe like I've seen a lot of Whitcomb's going around. My first question would be like, do you guys think it's necessary maybe to do to do a boot camp to.

[00:35:49] So you see, you have a masters in mathematics.

[00:35:53] Yes. But it's more in the theoretical, very like pure side of it.

[00:35:58] Yeah. And so, I mean, Denia boot camp, probably not. You can easily pick up any any book, you know, whether it's introduction to machine learning with Python or Hands on machine learning the python. Tensorflow and working your way through that book because simultaneously you'll get exposed to the could learning API, Tensorflow, API, Pendas, API, the programing languages that you'll need. And you also get a high level overview and understanding of how some of the commonly used machine learning algorithms actually work and are applied to solve particular problems.

[00:36:36] But, you know, if you do want to take bootcamps to recommend Dave's, it's a good one. But but bootcamps know I would focus like because how much knowledge you can right. At some point you have to use your hands and have to be tangible like I could. I can, like, watch somebody build how this has happened before.

[00:36:59] I've been on Habitat for Humanity house build and I sucked at it. I would never trust me building a house just because I know how to hold a hammer and that I'm supposed to take the hammer and nail. It doesn't mean that when I go do it for the first time, I'm not going to smash my finger against the wall. Right. You have to practice. You have to develop that skill. So take it theoretical knowledge and find ways to apply it. And again, and guessing like a broken record here. But it's all about projects, all about practical knowledge. I'll let somebody else chime in as well. Actually, my wife said it's annoying because I keep saying chime in. So, Romi, I know you're listening now because she listens to this all the time. I'm sorry. I keep saying chime in. I'll let somebody else contribute.

[00:37:42] I get asked this question quite a bit. Usually if I had to direct message on LinkedIn and in all of my career and I've been in the field for 20 something years now, so quite a long time. One of the things that you're going to need to study first I would recommend is applied computer science. You got to get a code. You got to learn how to write code. And I would start with SQL SQL. It is it is the lingua franca of Data. You can't go wrong. And it's not a waste of time. It's the first programing language that you're learning. That's great. It's awesome. Sequel's immensely useful. It's very approachable. I don't know what your math background is, but if you have any set theory, it'll be really easy for you to pick it up and understand the basics of working with sets and SQL and then move on to there and pick Python. Ah, after you've done some research on what kind of jobs in which kind of companies you want to work for.

[00:38:30] So I'm actually trying to learn SQL right now. Do you have any advice on which resources would be good for SQL queries for mere mortals.

[00:38:39] That's a good one. Or DBS course on SQL. That's another thing. So yes. Yeah, yeah. If doing taking the LinkedIn course and I think it's Black Friday, so you might even get a discount if you son of a day of course.

[00:38:53] But yeah that's the school stuff is free by the way. That's on YouTube. I'll put the playlist in there in the chat.

[00:38:58] Oh right on. There you go. Yes, that's good. And another book that I always recommend my mentees is SQL Queries for Mealer Mere Mortals. I forgot who wrote It's John's with the V John V something. I'll type into the chat soon as I get a chance. The more I can send you a link on LinkedIn for it.

[00:39:19] Somebody just entered Mark. What up. How's it going in Machree. How's it going Jennifer. How's it going. Sorry Chin if you guys Adam guys have questions let me know. Anybody who doesn't have the camera they turn like a black box. I just automatically ignore them. So I'm sorry. If you guys have questions just just feel it and meet yourself and go for it.

[00:39:41] Yes, go ahead.

[00:39:44] Oh, I was just going to add to Jacklin, if you wanted, like a short tutorial, like kind of like sandbox WS three schools is a good resource, but I go to and I can write that in the textbooks to Harp for just a quick add in all these kind of questions.

[00:40:02] Let's just say you're learning. New or you're reviewing something called minimum three references and you get stuck. That really helps to read it from multiple perspectives and like Harpreet Sahota and Harp billion in a good way dove into the concepts each time you have to go review something, use it again, try to understand the concept. By every chance I get to speak to my business buddies, I remind them, guys, all machine learning problems are math problems. Eighty percent of my work is turning our Data into numbers and making sure it's formatted well. And you get good at that. You get good at really connecting those concepts to the problem you're solving. It's not that it gets easy. It just gets easier when you think begin to think that way about each project.

[00:40:55] It's great advice, Tom. Thank you. So let's see, Jennifer, if you have a question, I'm just going to start calling people's names just to make sure that nobody's left unanswered. But, Jennifer, you've been so nice and quietly for a while. I've got a question. Feel for to yourself and go for it.

[00:41:08] So I do I am very fortunate that I work at a large company. We have lots of databases, a disturbing number of databases. So I'd like to get my management board into automating data integration and consistent dashboards. Do you have any specific suggestions for what goes into the elevator pitch to management over?

[00:41:37] Nicholas already answered this, guys. It's it's beautiful. I think he and I are of the same opinion here.

[00:41:44] You take those people out to coffee or whatever, something social. You get to know them. You find out what are your fears and how can I bring Data to your rescue and you get good at doing that. I, I don't mean this mainly about people that it's just you have to have good business acumen to be a good data scientist. I think that's barking up the wrong tree. You go and talk to your business counterparts like you have business acumen equal to them. They're going to shut you down, have a spirit of I'm here to learn what your needs are. Let me just bring a little bit of Data to help and then do it really small bit. It's like just, hey, I dug up the state and came up with this paragraph. Is this moving the right direction? Then over time you can move to some automated, beautiful dashboard, but start small, make friends. And Nicholas Lardy really said that above in the chat, so I want to give him credit for that.

[00:42:44] Yeah. So I don't know how much how many people know this, but I self described myself as a recovering enterprise architect. So this kind of thing used to be what I did for a living, which is why I don't do it anymore. By the way, to give you some sense of scale, my last enterprise architecture project was to create an enterprise information model for all of Microsoft, which broke me, by the way. So what I would advise Jennifer is start small. If you're going to try and take an enterprise first big bang approach, more than likely you are in my experience anyway. You are not going to be successful, that kind of stuff. The Democrat, if you know what team is, they haven't they haven't really made much progress in two decades for a reason. It's really, really tough to do what you're describing. So aim small, try to find synergies where you can ideally across organizations, if you can find two executives that are willing to cooperate for mutual benefit and if you can get a win there, you can then use that to start rolling that snowball down the mountain side and get a bigger and bigger. So team small Finta executives are willing to work together and start there. That's would be my advice. Hard one. Painful advice.

[00:43:53] One thing one thing I want to add quickly, Jennifer, is that yes, start small but as you grow, document, document, document. Because one thing that I suffer from, like I always see companies do, they have multiple databases, multiple clusters, import documentation that ties these databases to business processes. So as you are establishing that synergy, as you're taking on the small use cases and growing, make sure to document it so the big picture will make sense and adoption will be seamless.

[00:44:29] So, Genta, hopefully that answered your question does. Thank you, guys. Awesome. So let's see Jinyang. Have you had a question? And Markazi, I'll get to know these other people. I have a question. Jinyang if you have questions, feel free to ask yourself. If not, then I'll go over to Adam. So Adam, if you had a question, feel free to hop on the call.

[00:44:53] Like Adam asked the question what he did.

[00:44:56] Oh, shit. OK, you have experience writing complex securities, but not much with Etel and. The base design or administration, are there any recommended and recommended resources specifically for that? I'm going to bring that over to Dave because he just talked about his life as a recovering database architect.

[00:45:18] Yeah. So the first question would be oil, TPE or data warehousing, because there are two two different information architectures. Completely so. OK, so for example, I'm I've built LTP databases from the scratch way back in the day and spent a lot of time building data warehouses. So, for example, if you're more interested in the analytic space w how you build Data relational databases in a optimized way for conducting analytical queries and analytical processing, you want to look at the data warehousing realm and the guy that you want to check out is a gentleman by the name of Ralph Kimball. I'll put some stuff in the chat. Ralph Kimball is the man is the originator of the dimensional model, and he talks a lot about how you not only structure the physical tables themselves in terms of being optimal for analytics, but also he talks about the nature of etal. What is your pipelines look like, what a staging look like. And by the way, don't. And I think Joe will back me up on this. Don't think any of this stuff is old. School doesn't apply. It still does. Even today, even in the era of big data, most of the basics still apply.

[00:46:23] I think what I work basically full time is doing data engineering, data architecture at this point.

[00:46:28] Right. So I called the pipelining and Data Data, one of them is Drouet work on a lot. That's more like a real time data store. And I think that to add to what they saying to others, there's also right now, I would say real time and event driven systems are becoming much more, I would say, use across industries. And I would highly suggest, as well as you're looking at architectures, databases of pipelines, incorporate some sort of learning about real time into that. It's rapidly evolving field. And there's I think it's exciting. So you call the real time analytics or event driven systems or real time machine learning a second.

[00:47:16] All of that Kimballs dimensional modeling modeling book is awesome.

[00:47:21] Martyn's what is a data warehouse design?

[00:47:23] I think it was Kimballs book the classic.

[00:47:28] Yeah, I'll pop a link to Amazon here. The chart. I'm looking it up right now Mark.

[00:47:32] Oh, another company could be designing Data intensive applications from equipment for the weekend. That's awesome. Thank you.

[00:47:39] I was taking notes because that's exactly another problem I'm working on right now. So that's that's perfect timing. Another question I have is I just got on last week a really cool text analytics project at my company to kind of expand our product, to include that for for our surveys. And one interesting twist that wasn't aware of, which is there's two components, is, one, how do we what's the optimal wanky miles to be one? So it's going to be really simple, optimum way to fly, kind of bad comment. So maybe employees like trashing their manager, saying things that are inappropriate. Right. We don't want to flag those and bring us to the top of the ranks for their employer to see. And then the other options. We have international clients, so we have different languages we're working on. And so, like, how do you have to take into account both bad context and then also different languages?

[00:48:37] Analytics have labeled data and all this just unlabeled. It's all on label. It's all unlabeled. Do you have resources like a budget to use an external API?

[00:48:49] I know. And so I think I think the way that they're thinking for this be one is essentially just doing workouts. And Westmorland rejects this as a V one and not they're expecting some fancy model as more so to do a proof of concept to prove all right is worthwhile to go into the machine learning, rootin and get that label Data.

[00:49:12] Do you have specific words you're looking forward to, to immediately classify a sentence or document as negative or bad? And if anybody else has anything to chime in, definitely go.

[00:49:24] Yeah, yeah. So to give an idea of the company I work at, we work with employee H.R. Data and so we have Glassdoor reviews that we pulled. And then also we have the Enron data set that has a label component as well that has similar language that that's kind of about. And thankfully, I'm not the only like Data person on this team. The person leading this project is really well versed in NLP. And so she definitely has a way around it. But I have my meeting coming up next week and just try to be more informed before I go on the conversation.

[00:49:58] This this might be a little bit old school, but if you don't have a label Data. You could always Mechanical Turk it, that is actually a strategy I've seen and successfully implemented before, you literally just go and outsource it to a bunch of human beings that know English, assuming English were the language in question and have them go through and label the Data for you. And typically from I don't know where you're located, Mark, I'm going to assume the United States, relatively speaking, it's inexpensive because you're typically using people in other countries that work at a lower rate relative to the prevailing wage in the U.S. and that's one way around it. Now, of course, unfortunately, you have to run like a project because you have to, like, send out samples, validate how accurate those folks are. And if you find enough folks or a vendor that can help you, that is one way to start getting labeled Data. And that helps out a lot. It's kind of it's kind of old school, but it still works.

[00:50:48] Yeah, no, thankfully, we have a whole bunch like psychologists on my team who are very familiar with Amtrak, so I can definitely chat to them a little bit and see how that how they go about. That's a great idea.

[00:51:00] Anybody else familiar with this problem statement? I would like to contribute the crickets as they know. Great question, mark. Hopefully you got something out of that, if not little circle back for sure. But I going to take a second to make sure that either he Machree or Pusha me if he has got questions. Definitely go for it. Feel free to unmuted and take over. Christian has a question. It's somebody I meet themselves just now. I said I said you have a question.

[00:51:29] I have a question. So being a deterrent isn't how much you care.

[00:51:35] How much big Data knowledge. Yes. I mean, that's the great thing about this.

[00:51:42] So do we know like we need to know the big data? I mean, how does it work or is to something or it's not.

[00:51:52] So I'm assuming your question is which big data stack should I know and at which level of technical depth should I know? The technology stocks? I mean, be familiar with them. Right. And I think the thing about being a data scientist is your ability to be able to quickly come up to speed on any technology when you need it. Right. So right now, you might not have any need for any big data tech stocks or anything right now. Right. That's cool. But you need to be able to quickly get up to speed on that when you need to use it. So I think the ability to be resourceful enough to quickly learn a skill is more important than just having it right off the stack to what extent you need to to know it. I mean, what kind of company you trying to work for? Where are you trying to go with your industry like or do you have any clearly trying to work in as a data scientist working with Iot devices? Because Shechem, I need to know that stuff inside out. Are you trying to be a data scientist working at maybe a marketing company, maybe not as much. Right. So depends on where it is you're trying to go. So once it's already been done, it all depends on the industry that you're trying to go on to. So get clarity on that, I'd say.

[00:53:02] And also you want to work to all at once. No, I would say like the back in the day, like in a late 20s, early 2010s, I think it was a lot more important to know big data infrastructure, right? It was in the days like I do don't think around like 2013, 2014, Slark and so forth.

[00:53:23] Nowadays, when you interface with these systems, I think it's like less and less important to understand as search for data scientists like how SPARC works. Right. Because all you're doing is you're basically going to be interfacing through Zeppelin or Jupiter anyway. And so it kind of doesn't matter what it's going to do or just like spread out across nodes and stuff. So back in my day when I was doing a lot of that kind of stuff, it was, yeah, you had to know how the entire infrastructure work, but think you have to get what I could bring you up to. I think with more important things, and especially nowadays with even more, I would say the more modern big data infrastructure is actually it's not SPARC anymore. I mean, everything is moving back to the data warehouse right now. You're talking about data warehouses and Amazon that's going to be redshifts Snowflake and Google Bakeware and azurite schnaps. I mean, everyone we see right now, this is a lot of companies at this point. We work with all the companies mentioned, but it's like you're seeing it makes move back to data warehouses for anything you data. And this includes the ability to use Pythonic against this system. So, again, the amount of fallacy to have brought back into the structure, I think is is declining, actually, which again, frees you up to focus more on what you theoretically might be good at, which is analytics or machine learning or stuff like that.

[00:54:39] So what do you think is causing the shift back to the data warehouse?

[00:54:42] Is it because now I have a bunch of maybe the separation storage compute? I think that's a big driver right now. Storage becomes so cheap you don't need to have a new cluster anymore. Like the the notion of setting up an HDFC cluster and holding it together. I think that's that's long gone. It's like I was actually talking to a Hadoop engineer a couple of days ago about the. We certainly felt like it's like, yeah, I think in a couple of years you're going to have a good talk about this, you're going to agree with me, and you certainly agree that they're looking at moving off of her system and higher than just simply because it's there's this less thing to maintain. I mean, you got a full time job meeting. I do cluster. But that's like usual for basically no one these days if you can get away from it.

[00:55:25] So or imagine. I'm sorry. Go ahead, Greg. So sorry. I should you could do like me. Right.

[00:55:31] Who's on the business side and who's also not a scientist is no high level enough. I think if you focus on understanding how the data was collected. Right. How the data was collected and why power is stored, prepared to be processed from a data warehouse perspective, how is distributed and how you can pull from that to perform some analytics like Joe was saying. And then what other tools you have on hand to present the data as well. So a high level nor enough to be dangerous and also learn as fast as you can when the business cases show up for themselves, as Arpit was saying earlier.

[00:56:11] So principles, if what I see a lot of trends going right now in the digital space is actually it's a way from data processing and it's more towards Data the data governance data quality. Right. So I think the Greg's point you're going to center on something like if you have to know how SPARC works, great. The number of companies that are using SPARC is actually dropping every single day, to be honest. But the question that never goes away, it's like where does this data come from and is it any good?

[00:56:35] Right. Because the data is crap and good luck.

[00:56:38] So I don't think so. There's a famous saying in computer science that goes along the lines of a layer of indirection, solves an extra layer of indirection, solves every problem. And that's true in analytics because it is a technical pursuit. So you have to ask yourself why, if they want me to be way low in the stack and all this stuff, why do I need to know that if my purpose is to analyze data, provide insights, to drive the business forward, why do I need to know all this big data infrastructure stuff? That's what I'd be. That's what I'd be wondering.

[00:57:08] So hopefully I answered a question. She had a question about the Bible for reading protection quality code in Python, Hitchhiker's Guide to Python. That's what I used to see. If somebody had a question, was it was it pushing me or them?

[00:57:25] Yeah, yeah, yeah, yeah. So high. I mean, this is just an amazing experience. It's my first time here and. Well, yeah. So thanks for doing this. And you can call me Bayati. I mean I know my name is a little difficult, but yeah. I mean you can call me, you can call me by name.

[00:57:47] So my question is that I mean that can be enough emphasis on domain based, I mean domain based, domain based project products. We need to learn that. So I love solving a sports problem. So, I mean, I just want to predict what is the next big thing, except that we'll see what it's a slider. So in baseball. So I'm going to do things I started working on. I mean, as I said, I started working on the effect of momentum in sports. So I just want to know why a team goes for a twenty four or another or ten out of five or not. I mean, that's amazing. I mean, you can you can pull off certain things here. So my question is there are a lot of data sets around. So what do say I try to get certain datasets from Kaggle and I started doing the statistics.

[00:58:48] I ran the runs test to predict whether it's it's the weather, the winds are random or I mean something like that.

[00:58:57] But my question is that how do you how do I choose the right data and how do I choose? I mean, there are multiple ways to solve this problem.

[00:59:07] And I mean, is that I mean, is that a so-called. Right.

[00:59:11] So, I mean, there's no one right. Where there might be ways that make more sense in other ways. But fundamentally, I don't think there's a right answer to anything in Data science. Why? Because we're just using our simplest comprehension of reality, which is these mathematical formulas, which are just ways to describe the natural world to solve problems. I want to get a philosophical debate here, but I don't think mathematical statistics is real that are just Data we invented them as a way for us to try to understand something in general. There's never going to be a absolute right way to answer any data science question that's just did not want to get to philosophy, my philosophical view on that. But that's just how I feel about that. But I would recommend Kenji's sports statistics, the YouTube channel. Um, maybe I can get them on. To show one day, but check out his sports analytics thing and, um, definitely get some insight from there in terms of where to find Data, I'm pretty sure most sports records are available publicly. You might have to scrape the data yourself and source it yourself, but it is publicly available. So go to whichever league it is that you're interested in. I mean, I'm just going to assume it's cricket. Sorry, I don't mean to be racist. I mean, you do is I'm assuming it's cricket, but yeah, you can go to their website and whatever the National Cricket League and download all the data there or scrape the Data. So yeah. So it's a lot of funny looks. And Pimas told people that statistics and mathematics isn't real. Jacklin Sorry.

[01:00:49] Can I ask can I ask a question here for the professional data scientist, especially the big teams or school systems into production?

[01:00:59] So I'm on a yes journey to raise awareness on.

[01:01:08] You know, we we see all the hype about, you know, intelligence email and we want to replace this with automation. In one analysis, especially people like me forget to do is the cost benefit analysis. And, you know, yeah, I want to automate it, but it's as expensive as the manual process. So you're really not solving anything kind of thing. If I'm on the business side, what do I need to look into inside of the production process to key in maybe a unit cost of that production to justify or reject that automation? You know, so, you know, this is something I definitely want to raise awareness or cross program managers. Product managers don't don't fall for the hype yet. Make sure you understand first. It's not just it's not a rule-based problem that you're going through. It takes more than no base. Then it's repeated Berlanti error, etc.. So next, let's move the machine learning. OK, now you move to machine learning. There's a system already in production. Is it worth it? Yes, I know. So you can help me pinpoint where to look to kind of put a cost behind that automated implementation.

[01:02:27] There's got to be some setup costs, right? So you have your fixed cost. Right. And then on top of your fixed cost, you can add whatever variable cost there is to actually do the actual work. Right. So there can be some Aggregative, whoever is going to be working on it, times the number of man hours expectorate so that the input cost to making this thing automated, you can estimate that. Great. Now it's let's think forward. How much time are we going to save by implementing this automation? All right, cool. So how many people are going to be using this automation? How many hours is it going to be this way?

[01:03:00] If one person uses automation, how many hours is it going to save that one person on average, that one person whose time gets saved? How much is that one person's hour worth to the company? Multiply those two together, multiply that by fifty two weeks in a year. That's how much that one person is saving and cost over the course of the year. Multiply that by the number of people using the automation and now you've got a estimate for the amount of money you're saving the company.

[01:03:25] Now the setup cost, the estimate set setup cost, you know, cost benefit analysis there. So that's how to approach it left to everybody else.

[01:03:34] I see the Greg, you're speaking my language, man. OK, this is this is what I used to do for years. I was coming up with I.T. projects, I.T. programs in game kit, getting them sold and then delivered. So the good news is in a very real way, building a mall based production system from a cost perspective is no different than any other IT project. Generally speaking, all the things that Harpreet talked about totally fit in. Here's one thing that I would very, very much recommend now. Find a business executive that's willing to commit in writing to the cost of benefits, whether it's a higher sales or lower costs, whatever the Arawa is the company. Make sure you've got a VP, a director, somebody on the hook that says, yes, if you build this, I will commit that this is the value that it's going to bring. Believe it or not, oftentimes that might kill your project right from the get go because nobody wants to do that. So get that. If you get that, then you say, OK, cool, you can price out the infrastructure, working with the engineering teams, all that kind of stuff. The one gets you the one Gocha that I've found in email systems and I've been bit by this by myself, is accurately forecasting the cost of retraining the model periodically through the life cycle of the system. That's the hard part to get someone to nail that down. And that actually tends to inflate the overall TCO, the total cost of ownership. And oftentimes small based systems, at least in my experience these days, often do not pencil out because of that, because the benefit that someone's willing to commit to. Once you start saying, look, the model needs to be investigated every quarter, every month, whatever it is, and possibly retrained, new features need to be engineered, all that kind of stuff when you had that cost and sometimes it doesn't pencil out. So that last bit is critical. If I'm going to refresh the model, how often how much is it going to cost?

[01:05:20] Each time I refresh it was actually I was actually writing about this this morning because it was a notion of total cost of ownership. Right. But there's also the notion that we come across the total opportunity cost of ownership, which is the notion that if you're going to make an investment in an initiative, whether it's an island automation or whatever, there's obviously the total cost of ownership considerations.

[01:05:40] Right. Operational costs, whatever a cap expenditures need to put into it. Right. And that ongoing operational burden of doing this nowadays, what I would say one consideration is really have we think about is especially you working with engineering teams, what is the preference of technologies the engineering team may have? Are they going to want to involve themselves at some standing up open source software? You want to use a managed system or proprietary because what you want to work out is like what do you have options in reversing this decision if you get stuck it right or are you stuck with this infrastructure now that you're going to be maintaining for 10 years plus, in which case it's not just the models that you have to worry about. It's also the underlying technology that you just built and inherited it for yourself. It's a really big thing because now you have a team dedicated to this. And if you don't have optionality, it's going to be really hard for you to switch to new projects because your team is succoring this. We see this all the time. I can't tell you how many Hadoop systems are trying to move people off of right now. They're like, oh, this is like the Harp thing in twenty ten or twenty eight or whatever. Let's obviously going on this network. Dear God, how do we get off this thing. This is expensive and it's creaky and it's falling apart and our machine learning, suffering and everything else is suffering. So it's like we've got to consider like what's your outlook as well, like activity versus decision or I stuck here.

[01:07:01] So yeah. So the speed that I'm seeing this, you know, these technologies are evolving. You know, I'm assuming the more you have new technology come out, the more expensive it is to retrain the model throughout the production lifecycle. And, you know, I'm assuming that that future cost of retraining is expensive, like they were saying, maybe get it right, inflate your TCO. And how do you pinpoint that that future cost that that's one thing that's in the gray area for me. So I do appreciate the the insights here. So thank you for answering my question.

[01:07:39] Thank you for asking me. It is a great question, a question here from a Machree philosophical with it. I see this a lot. We need a good data scientist. What's the fine line data scientist and a good eye.

[01:07:52] I can confidently say I spend zero percent of my time thinking about how to be a good data scientist. I focus most of my time thinking about how I could be the best data scientist that I could possibly become, and that is just focusing on making sure that I learn more than I knew the day before, developing my craft, honing a sense of mastery for my craft.

[01:08:13] And actually, this is a common question. I asked a lot of people on my podcast, a lot of the Data science leaders. I'll ask them what's the difference between a good data scientist and a great data scientist? So I would say don't worry about how to be a good data scientist, worry about how to be a great data scientist, and I will open it up to whoever else wants to answer that question. Go for it.

[01:08:34] I would like to say something that's interesting.

[01:08:38] Thank you, Harp, for these sessions. Those are amazing.

[01:08:41] I'm really not allowed for nor do they have some takeaways.

[01:08:46] My takeaway that I would like to share with you, the first thing is to have the capacity of learning fast and learning by doing this.

[01:08:57] I think this is the best thing that is very important.

[01:09:01] If we want to become a good design is another thing is to be creative. And oh my God, this year is is we creative to propose creative solutions? Because when in your view, you need to do some projects, for example, that is expecting to do that.

[01:09:30] And I want to tell you, Jaclyn, that I would like to have all the knowledge that you have. So you are your starting point is like top. So I'm sure that you are going to be an excellent at the scientist. Thank you.

[01:09:45] That's that's encouraging.

[01:09:48] Yeah, I have it. That's how you become a good data scientist. Sorry, I didn't mean to cut you. I'm gonna go for it. Go for it.

[01:09:54] Oh yeah. Yes, yes. To to finish this idea is another another thing that I have not ne.

[01:10:02] In the interview was is that when they ask, present yourself, talk about yourself, you are presenting Data, you are forsaking your own Data and this is the way they can see your communication skills, because if you are willing to communicate, the results of all your projects are all the dash for the things that you have produced with a Data at the stakeholders, you are not able to present yourself. How can you present the results? So it's very important the way that you present yourself, because they are in that question. They have seen a lot of information and they I think the key is how you present Data your own Data that you know everything about yourself, how you present a person who has to decide if you are the one for that thing. So I think these are my ideas I want to share with you.

[01:11:11] What do you think? They want to hear what you have to say.

[01:11:16] He is quite talkative earlier that he has been a great data scientist.

[01:11:24] I guys. Well, we are over time. Thank you so much for coming out and hanging out. Everyone apologize if I did not get to questions, but I will be back here again next week. And I hope you guys will as well keep an eye out for the podcast. We got a really cool episode coming on Monday with Nir Bashan. He wrote The Creators Mindset, probably one of the top ten books that I read this year. So check that out. I really enjoyed it. And you guys have a good rest of the weekend and good kick off to the holiday season and I hope to see you guys next week.

[01:11:58] Take care, everyone, I think, etc.. All right. All right.