Fireside chat with Eric Weber - Learnings in Data Science

Key Takeaways

hello everybody welcome to yet another mantissa designs webinar slash fireside chat today we are talking with eric eric weber i don't want need to introduce you to him but i'll just give a short introduction and allow him to introduce more in detail eric weber right now i think is senior director and head of data insights at list reports uh and he's he's been a prolific writer on linkedin and that's where we've interacted via text but this is the first time we'll be chatting uh with each other face to face so i'm very excited for that uh and yeah he was earlier at linkedin and he's been in at many different companies so without further ado let me just uh get eric uh on to the on to the discussion uh eric thank you so much for joining us today uh i tried to give a good enough a basic introduction of you but i don't know if i've done a good job maybe you can start off with an introduction of yourself you did a great job thank you um as far as intros go um i never know exactly what what to talk about but i think the key the key ideas are i did start writing when i was at linkedin um back in 2017 and it's kind of remarkable so a lot of people that you see active on linkedin like fabio vazquez and um kate strashney and everybody else we all kind of started riding around the same time and kind of awesome to see where it stands just a few years later we're having we're having fun with it um it definitely has like posting publicly has its moments um but the data science community and not just the community in the us but the community worldwide is really strong and that's something that you're seeing with having so many people active um as far as um you know me personally i'm from the academic world originally so i was i used to teach statistics and experimental design i got a little bit tired of how long it took to do anything in the academic world like a paper would go from um you know you'd submit it i submitted a paper in 2009 and it was published in 2015. like you can imagine how like impact is slower so i prefer the business side getting to go quickly to impact is a big deal for me and i've been in different i've been in different roles along the way um and yeah so this is actually an interesting times like yesterday was my last day at list reports and monday is my first day at yelp so we are moving um it'll be interesting i'll post more about it on monday specifically about the role but again it's going to focus heavily on um experimental design heavily on analytics which kind of goes along with a lot of the things that i write about and talk about um that's something to note most of the things that i talk about are generated from things that i experience every day right it's a that's the best content you can generate is thinking about the issues that you come across yourself um because it comes it's much more authentic and it's meaningful to people so and we can start there and jump into however you want to drive the conversation sure uh what's interesting to me is that you mentioned you were in academia uh for some time were you a researcher did you do a phd yeah so i did i did my phd in math um with a heavy focus and statistics so the academic world i was a 10-year track i was on the tenure track as an assistant professor i was at oregon state and then at university of minnesota but five years in i kind of i don't even know how to describe it i love the teaching component the research component was fine but when you think about how you're evaluated in the academic world it's very much about how many articles did you publish how much impact did you have i think it's kind of ironic that for all of the articles that i published in journals a single post that i make probably get read by more people than all of those articles combined so it's kind of it's kind of a it's it's really interesting to think about that like how much i put years into writing articles and then i can write something in five to ten minutes and i get seen by more people so i think that's something to consider too like the reach of what we do and podcasts posting you've seen it happen with your own account right you can get you can reach a remarkable number of people um and people are much more willing to read a short form post then they are willing to read an entire journal article i think that's that's in general with uh the changing attention spans of people around the world right there's rise of apps like tick tock and all the short form content apps so uh no so i the the reason i brought up phd i'm still not getting into the template that i shared with you the reason i brought up psd is because there's a lot of questions around do i need to be a phd to get into data science do i really need to have that background so you had that background but how would you answer this question um no it's a it's a pretty simple answer um a phd look at it this way if i see that someone has a phd i think about it in this manner potentially they have a really specialized skill set so they would be apt for like a research team at microsoft or google or something like that like for those teams yes a phd is generally necessary because you are literally doing research you're you're basically asked to publish a prominent conferences around the world and so in that case sure but 99.99 of positions and data are not on those teams a phd to me if you have it signals a couple of things one it's really it means that you went through a lot of pain typically like it's it's honestly a signal that you can work extremely hard that you can work on complicated ideas and get something done it's also i think it's also a it's a nice to have but any company that any company that says we must have a phd you must have a phd to be in this role typically that means to me that they don't really understand what they're looking for just like going to a prominent school carries away with it having a phd can carry weight with it too but it's not necessary for entry um and i don't think it really ever will be because if you think about the time investment for yourself a phd versus i mean i do encourage master's degrees like i think if people can find a way to make it work financially and time-wise master's degrees are a very positive addition but a phd is like tacking on three more years and at that point if you think about just the time and like you could spend three years learning on the job rather than writing a dissertation and i think that's important but i do see that conversation and me i'm skeptical of any company that says you must have a phd to do this job like it doesn't make sense to me sure yeah in fact i think the conversation uh i i try to read a lot of online stuff and i think there's a lot of conversation on how you need to have a phd to be able to work in this field and all that but that is also u.s centric from what my point of view i don't see a lot of that conversation happening in india because we i don't see that like that as an expectation for you to be having like it's okay if you don't have a phd in india so that that's a divide in terms of the market itself i i'd say i would see that it's not only it's not only u.s centric it's silicon valley-centric like yeah so i think it's important to keep in mind that a lot of the advice you hear and descriptions of interviews and processes and stuff like that is very often based on its bias toward larger companies like if you think about um google and facebook and companies like that they tend to yeah people can people tend to focus on what's going on with them but they're not representative of the industry sure and that is and so yes especially outside of the us like i don't even think having the conversation do you need a phd is useful because the resounding evidence just says you don't yeah okay so maybe we can get into the template so one of the questions i had was what do you do in your current role which i don't know if that is a question i can still ask so i can tell you what i'm going to be doing i guess um going forward so the group at yelp when you think about experimentation right on a large platform what it means to run you know hundreds of experiments simultaneously it requires um engineering it requires data it requires building internal products so that people can run and interpret these tests so in a lot of ways it will be a partnership between engineering data ux design product managers things like that so it's going to focus heavily on experimentation and then i'm also going to be focusing on trying to you so yelp's data is very much focused on small businesses um and so with covet and with the economic changes the impact on small businesses is really important for understanding economic recovery so we'll hopefully be doing things that are a bit more public facing when it comes to sharing about on perspectives about the economy um so yeah it'll be fun i mean it's interesting yeah yeah yesterday was last day monday is day one so again that could completely change but um but in general in any role that i'm in it's a combination of you know like technical leadership but also managing up and managing down right like if you think about data science leadership you have to be able to lead your team right and have them trust in your ability to do that and help them think through problems but also you have to manage expectations absolutely right like what is your team doing why am i spending this much money on your team like all of these sort of issues come up um so that will be the day to day um that's and yeah it's interesting yeah that question kind of fits right in the middle but um it will be fun over time i'll be able to share more about what's happening sure sure so it's interesting uh i think you were doing some experimentation work in in your previous role now and now at yelp as well you would uh be continuing that but a lot of people uh who tend to traditionally talk about data science they don't talk a lot about experimentation as part of data science like i mean it's it's seen as part of data science but it's not seen as a primary part right like the focus is more towards statistical learning or the machine learning methodologies rather than so have you always been in this area is this something that you've developed interest on lately how do you see this um it's sort of been an interest of mine since i was in the academic world like experimentation was something that i've studied a lot and worked with a lot and i think it's important to consider the value that it provides like it's not traditionally viewed as the core of data science but if you want your model to ever make it to production or impact um or you want to change features and do meaningful things you're going to have to understand its impact before it goes live right and i actually think experimentation is in a lot of cases what keeps models from ever getting into a business context is because they don't understand what matters or what impact it's going to have because the first question that people ask before you make a massive change is what's going to happen to user traffic what's going to happen to this metric and if you don't if you're not able to tell them they're not going to approve the change yep yep you need to have all the answers before you can propose something yeah you do you have to have and it's not just a something that i would encourage people to think about as well as data science is a is very much like a product focused enterprise like typically you were trying to help product make better decisions about what to do and so part of that is doing an experimentation it is it's like uh and and often and this gets back to something i've posted about many times right like generally people coming into data science are too weak in statistics that is like a core it's a core issue because you're limited by what you can do if you don't have a good understanding and this doesn't mean that you need to have an advanced degree in like all the areas of stats but you need to have the core foundations really well mastered to be as effective as you can sure um yeah so i think that leads really well into the next question on how did you actually get started in the field i i know you've mentioned a little bit on your phd but maybe you can just give us a broader outlook on that yeah this is actually an interesting story so in 20 let's see 2013 2014 i was kind of getting frustrated with academia i was like i had just submitted i had just gotten an annual review and they were like well you you published four papers but you should have published five and i was like this is like yeah i can't deal with this for much longer um so i was actually talking with my dad and he was an engineer at um ibm at the time and i posted about this before but he like at the time ibm was starting to do a lot of their work with watson and stuff like that and he was like well like you should check out like big data stuff and like uh-huh sure because at the time i was dealing with i was dealing with relatively small scale data sets i wasn't having to think about distributed computing or anything like that um and so i did and i was like wow this is cool and honestly the first the first formal data science course that i took was like with data science and the title was on coursera from johns hopkins that was in the 2013. so like this is also why i'm a pretty strong proponent of like it doesn't really matter where you start in the field you just have to you just have to keep moving once you're in it right it's like there's no perfect start there's no perfect resume thing that's going to get you a job like yeah i was taking that's why i have like my coursera certificates and data camp certificates on my profile because they're useful so from that point on it was trying to figure out how to make um how to make my interest and statistics and academic background useful for business okay and it's a big mindset change right like a lot like really have to go in and be instead of something like i'm going to work on this research project you walk in and you have to say now i need to help the business make money at the core of it that's what it's about right like you can we can talk all about like oh we want career growth we want to do good things for the world like if you're gonna do good things for the world within a business you also need to be benefiting them financially it's just like it's a hard truth people don't like to talk about it but that's how it is it's also because where we sit right if if you're working as a data scientist or data analyst you're sitting in a room of executives you're sitting with the people making the decisions because you're trying to help help them guide in their in their decisions right so uh contrast that with say an engineering team uh which is trying to build the platform it's not really tied to the business metric i mean it of course it is like but you're not monitoring business mix if you're in a name so i think that's all sorry i think you might be breaking up on your end hi eric i think amazing is facing some some net network issues well that's okay so yeah i mean i mean i'll i'll take it for a little bit and we'll see what he um this is also the adventure of everyone being at home it's really fun um so what he was speaking to is and i and i have questions in q a and we can run from there um but what are you speaking to is data is very different from engineering right it's a when when you're working in engineering space you can kind of work on your project without having a mind or attention set toward the business right you're just working on a feature you're working on some level of details and but the difference in data is that the stuff you do often translates directly to monetary impact money whatever it may be and so that mindset has to completely change it's something that's uncomfortable for people people that come from academia are and this is true not just in the us this is true in india this is um wherever you're coming from you have to be hyper focused on helping the business generate revenue or make better decisions if you aren't then they're going to wonder what the value is um of the work that you're doing so um there are some questions that i will start taking um and i'll and i'm gonna take these um and i'm gonna take these sort of in in order like to make it somewhat coherent with where we were starting um so what i'm and so there is a question about how about if you're not from an engineering field how do you get into data science right this is a fair question um generally speaking it like to me it doesn't necessarily matter if you're coming from an engineering field or not it's nice to have often because the technical skill sets that are present um are often helpful to give you a chance to really dig in and really get to understand the infrastructure how to work with data but generally speaking it matters more that you understand business and understand business problems so if you're an analyst if you work with financial data if you do anything for a business and you understand their problems that's a very fair way to get into data eventually getting into data science because the core of it is that you can ask questions that are intelligent that help the business make smart decisions whether you're from engineering or another field that doesn't necessarily matter but in terms of preparation the most important things you can do one you have to have a core foundation in statistics you absolutely have to if you don't have that core foundation and stats it's going to be extraordinarily difficult for you to make the progress into the field that you want to make um the second thing is to get comfortable working with data that is not not clean um and so by that i mean most data sets that you work with online can often be i'd say they're too clean like they're set up to do really clean analysis with in reality most of the day that you deal with on a daily basis is going to be messy it's going to be unpleasant and so finding data sets that actually give you a reasonable picture of what's happening in the real world is really important those are the two things if you're going to start statistics and also getting familiar working with data that is not super clean um so when i think about books to read in mlnai again i'm speaking from a from a business perspective here there's a lot of really good lists of actual machine learning books um what's happening is that you need to find books that focus on business so there's a really well known book called data science for business it does a really good job of discussing data science statistics techniques in the context of helping a business make decisions it's it's relatively well known um it's pretty prominent the author um actually unfortunately passed away recently and but the book was one of the first things i read in the field it's incredibly important and it helps you really get a sense for how do i talk about modeling and stats and data in the context of actually making money for a business so um i know we jumped on a couple of the questions so it looks like looks like you're back in business i'm sorry internet activity is always an adventure so um i just started taking some of the questions from q a so we're totally good to go great thank you so much for picking it up uh i don't know i don't remember what we're doing talking about let me oh good we can pick up wherever you want so okay okay so yeah you mentioned uh about how you got started uh it was your dad that helped you get started which is uh yeah like like how most indian parents are i don't know how how how it happened there but anyway i heard that story and um when i was in bangalore last year i was like like yeah my family helped like that's how it started for me too like okay there's a lot more similarities and differences i love that all right so that takes me to my next question which is a very interesting question what is something that you would do differently today if you were to start today so whatever you did when you started versus now i'd worry less about trying to learn every tool and instead focus more on understanding the core of the business that i'm in there's a temptation when you get into the space to learn all of the tools learn rn python learn sql but in reality it's the thing that takes the longest to learn is generally the business sure understanding understanding so if i look at someone who comes into a business and whether they're going to be successful yes technical skills matter but what matters more is that they are able to understand what metrics matter to the business right like a lot of what we do in data science is actually um is actually focused on metric development right like coming up with useful metrics is really hard it's really hard and it also requires um it also requires so much how do i put this it requires so much deep understanding of what matters to leaders what actually is useful because typically what happens is leadership will focus on typical metrics right like if you're in a if you're doing a product they'll focus on daily active users weekly active monthly active that type of thing but that probably doesn't tell the whole story um those are like vanity metrics right what they what you care about is engagement and so i would focus less like if i was to go back and tell myself in like 2014 like here's what's going to matter no one's going to care if you were doing it in r or on python or sql or excel they're going to care that you can actually figure out what matters and so it often involves like so i used to think about onboarding um new employees from the perspective of we have to teach them all the tools but it's a lot more about teaching them the business like the tools they can figure out along the way but like if they don't understand the business six months in then there's like there's no chance that they're going to be able to do anything useful sure um i think uh you made a very good point on the metrics and i see this also in my work and work of other people as to how you can align uh your technical metrics with the business metrics and again uh reconcile both of these so that whatever technical suggestions or technical decisions you take are aligned with the best for the business yeah so i think we often want to i mean a good example would be you could come up with the best model in the world and that's not necessarily going to be useful for the business in that moment like you could something i see often is people come up with models that end up saying like hey this variable is really important or this variable is really important fine but if you come into um if you come into a presentation and you say this is important or this is important and the variables you identify are not actually actionable right like essentially the business can't do anything about them then let's like think of what the miss you're gonna okay well the success of our business basically depends on something entirely outside of our control told them nothing that can help them make a decision internally sure you may have built an awesome model it doesn't matter like so those are good examples about making sure that what you build and create um is aligned to the business and align to the value that they need a very related question on that so how do you propose somebody uh who has no and like again a lot of people who are coming into this field are being said that like i said if you learn r python sql all of these things you are likely to do well in data science but uh hearing from you that is not what you're saying you're saying you need to have the domain expertise you need to really understand how yeah it's it's more about um so my question was my question was how do how do they develop the sense of understanding even before they come into the company so so for me my perspective is that you probably won't have the sense developed before you enter a company like i and so i don't want it to come off as i'm saying you have to have developed this expertise before you go into a role my point is that you have to have the mindset so like when you're interviewing right when you're interviewing and they ask you like what questions do you have for us right if you ask what do my first few months look like and how are you going to help me learn about the business so that i can deliver value like that will shock them because nobody ever asks that question nobody needs this i mean they almost always ask well what tools do you use what tech stack do you have um what models have you built recently yeah i'm guilty of this yeah yeah it's so interesting and so like just being but but it's on two sides it's understanding that as a candidate that it's important that you have you can go to a place that is going to allow you time to learn the business but also i think sometimes the business doesn't understand that they need to teach a lot right like you can't just bring somebody in and say okay read this paper and now you understand our business like if that's true then your business probably isn't very worthwhile right um i think you sort of answered the next question that i had so the next question was what are some of your learnings from your experiences so far so yeah i think most of the year yeah okay so what are your thoughts on the different roles in data science per se so what do you what do you have to say on that a data analyst versus a data scientist versus a business analyst do you see them as different roles do you see them as a kind of a fluid structure where everyone's flowing into each other what do you think in a perfect world there would be different roles but in reality they're not because titles don't make any sense like every company has its own rules about what they call data science business analysts data analysts and so for me i think it's more productive to think about like the title like data professional and then figuring out what is it that you do within the company that matters um i don't think we're ever going to get to a place where we have standardized titles across companies so like we shouldn't hope for that but what i do hope is that we start to understand that by defining a role as an analyst or a scientist it actually is helpful for everybody because it tells candidates what you're looking for it gives like you know we talk about data science a lot but some people they like doing the analyst work they don't want to do data science work but if you create an analyst role where you also expect them to be a data scientist yeah like that's challenging for people it's intimidating it actually leads you know what for a lot of people it leads to getting burned out um because they're suddenly asked to do all of these extra things and so for the roles i in an ideal place i see them as relatively distinct from each other and that analyst focuses much more on metric development business reporting trend analysis scientist focuses more on um model development thinking about scalability of models and insights but in yeah a few well-defined roles like the role of a data engineer is uh pretty well defined that way yes so data engineer is pretty well defined um i would that might be one of the few that's well defined i mean data like when you think about database developer data engineer like it's pretty well understood what they're going to be working on yeah it's like the data science area people just like hey go work on the data i'm like okay what does that mean um even something like a data science manager can be mean many different things right and some in some companies that means you're not going to be doing hands-on work and you are going to be just literally managing a team in other places it means that you're still going to be spending 60 or 70 of your time doing individual work and it depends on the needs of the company so i encourage people instead of focusing on the title focus on the job description like what what do they actually want you to be doing whether it's an analyst role or a scientist's role like that can often that can also often be negotiated um in the process like if you get through and they want to hire you and be like hey the work you want me to do is more of a data scientist role than yeah it's okay to push and this is true not just in the u.s this is this is true all over the world because a lot of companies they just don't know what they want right that's the core of the course i've seen this this very interesting uh trick some of the companies use as well so they would uh title your role as data analyst but they would ask you like i said to work on a data scientist position and and pay you less because uh that's that's the key yeah it's like all of a sudden you get paid less you're like you know yeah i was i was like it doesn't match up it doesn't add up like you are calling him a data analyst and making him work as a data scientist but you don't want to call him a data scientist because you don't want to pay him that much it's crazy that would be more in the side of ethics and hiring like but i do worry about it and i want people to know that like you know companies do that kind of stuff and you have to be you have to be careful to make sure that you are getting what you deserve um for the position that you're in because this happens and it's frustrating you mentioned burnout i actually made a note of bernard it wasn't my in my initial uh set of questions and also you wrote a post about it a couple of days ago on burnout and also it was a question asked by one of our listeners in the previous version of this podcast so i am seeing this theme re-emerging about people getting burnt out in this field again it's due to various reasons one of the reasons i see is because there is so much that you can never know if you know enough right so what are your thoughts on this um i think burnout in this field is not so different from burnout and in others if you look at like this is more u.s centric but like a very large portion like half of the teachers who start in the u.s the field within five years like that's been true for a long time so what they like thinking about why it happens is really important right there's a and it happens in teaching same way it happens in data people go into the field not really prepared for the day-to-day work and by that i mean there's the teaching for us there's the data um but there's so much else around it like you're dealing with people you're dealing with management you're dealing with people who are trying to come and ask you things all the time who don't actually where they treat you they're like go pull the data for me right like this that type of stuff is something that people don't expect and it and it weighs on them um and so when you think about burnout it's often like a misalignment of expectations nobody really told them what to expect and so they're like well i'm just going to be building models and doing this all day that's not actually what's going to happen i promise like but that's sold to you right yeah that's sold to you from the outside right from uh yes when you're not in the field i think so one of i'm not going to get into this too much because upgrade is our sponsor and they are an ethic company so i'm not going to get into about how companies education companies talk about this a lot but point is that yeah i think managing expectations and understanding what you are signing up for is uh really key there like you said uh it's very important yeah so next question yeah so yeah while you were talking about this is not there here as well you're talking about getting a data dump and uh asking to analyze so where are you uh on the receiving side of such a data dump any time in your career so far oh yeah i mean yeah probably like four days ago i mean these things happen i mean people assume that as you get further into your career that this will stop happening but it doesn't really stop happening and it's because when you go into a role and you um and you do good work people start to know that they can come to you for questions and insights so it's the thing good work creates more work usually like do a good job and then people are gonna be like well i'm gonna go to you every time i need something um but it's true and it and it can be tired because you want to help people out right like i don't want to tell people like no go away um sometimes if what they're asking for isn't aligned with the priorities of the company like this is where this is something i've had to develop at like a personal level just saying no right like i can't you can't put it in but it's hard when you're starting out in data you may be new into a role and you want to impress everybody so i think this actually goes back to this creates burnout because you start stressing yourself out by trying to do everything i was actually trying to come from it from the perspective of uh like like somebody comes to you and says like here here's a data dump now find insights for us right like that's a whole other problem yes i was actually trying to come from that perspective yeah so that happens too um so it's not just a data dump it's the what you're talking about is the the core of the issue is that people assume that because they have data there's something to be found yeah right like absolutely we have lots of data we should find something here i collected this data analyze it to tell me what's important but and that's an organizational issue often is you have to understand that the data you collect controls the the possible conclusions you're going to be able to make and you can't just you can't just jump in and say like okay now we have data there's going to be some magical conclusion that comes out the other side but it's true and so you have to manage expectations i've had to do and it's hard because people are i want a clear answer yeah but most of the time you don't get a clear answer from observational data so like you get at best a potential signal or conclusion and that's really hard because people i thought data and statistics was supposed to tell me the truth no no yeah yeah it's interesting yeah so i think and a lot of uh beginners fall into this trap of trying to find insights in all sorts of data right because now you've just started a new job and you want to be able to contribute to the company you want to be able to see it as a good worker and yeah now your boss comes to you and say hey here's a data dump can you find something for and then you start seeing correlations where you shouldn't be seeing correlations yeah right so yeah i think yeah so my point was uh has this happened to you and how do you deal with this i think you answered that sufficiently well uh i think now let's move on more to linkedin and what what it has done for you and in general how important do you think is networking on linkedin i'm i'm i mean again i'm going to give a biased answer because it's pretty much but the majority of people that i've met have been because of linkedin period like i mean i it's a bit unique because i worked there so it changed but i would have never posted on there if i didn't work so it's all this weird all this weird confluence of things but networking is critical um and that doesn't mean that you need to be posting frequently to get people following what you're saying it means that you need to create an image of quality and thinking and by that that could that doesn't have to be an original post um some of the most insightful things i've read on the platform come in comments and responses like when i post i try to start discussions i try to not i try to not create like hey this is the truth come get it i treat it like not like r sucks python is the best nothing yeah like it doesn't it's not useful for anybody yeah um so i try to start discussion and so if you think about networking it starts small like connect with people write a meaningful message to them interact with people in the comments um and things like that build over time like for me like i'm much more likely to end up connecting with somebody if i've interacted with them in the comments if we've talked before like building those relationships matter and and for me it's and for me it's kind of changed my entire honestly it's pretty much changed my entire life in a lot of ways um i mean i've gone to switzerland a couple of times i went to india last year i was supposed to go to africa this year to like like you don't really know where things are gonna go until you just start doing it and you have to do it regularly like you're not gonna post one viral post and like everything is gonna change as you know right you kind of have to just be consistent um and in terms of networking though yeah your network is gonna more or less determine where you go in your career so being strategic about how you build your network is critical so of course now you mentioned about how you try and start discussions uh well through your post but you you perhaps can't do that when you're just getting started writing on linkedin right so how did how did you get started writing on linkedin now i think the next question i had was uh how do you decide what to write on linkedin so i think i'm going to tie all of this how do you how did you get started writing on linkedin what what were your initial thoughts on now of course you can't start discussions when you're just getting started yep so exactly how how did you go about that um so i think the most meaningful i think we said this even at the outset um the most meaningful thing you can do is to write about your actual experiences that you're having day to day um trying to look at what's popular and then posting the same things as what's popular often won't get you the attraction or the start that you want like you need to look at what did you have trouble with what have you experienced in interviews job searching what people really want when you look when it comes down to it is they want authenticity right they want they want to know that like this person is really gets it they experience like challenges day to day right um you can contrast a couple of things one would be like there's certain posts that show on linkedin where it's like there's no chance that it's actually a real story that happened like there's no way like there's you see those come across like super popular posts and like there's zero percent chance that actually occurred but people want to be able to say hey what did you experience what did you struggle with so when i write about sql and i write about literally the frustrations i have sometimes and like data cleaning that kind of stuff resonates with people because they understand it like good and so i looked in that direction um thinking like okay what matters to you because what matters to you is probably going to matter a lot to other people um in terms of deciding what to post i wish i had a better like answer somebody i kind of really so i typically write my posts twice a week um i set time aside i don't write them brand new most times sometimes i'll write them spur of the moment but it's hard for me to just wake up and decide to post something i need to think through it and so when i think about those topics i tend to write so during my work week i have a sort of a notebook where i'll write things down as i come across ideas and then i'll sit down and write a few posts um two two different times per week and then i'll share those during the week um i think it's it's good to keep note of that because it's sometimes it's easy to forget your good ideas right you're like man this would be a really good idea and then suddenly you're pulled into five more hours of work and you forget so but when you think about ideas as well um more general ideas tend to resonate with people more than really really specific things so if i even though i am interested in nlp i probably won't post about it on linkedin i would post more about that on twitter probably where i know the people that care about nlp um linkedin is i would say a more generalist audience so like when i write about data professionals and things like that i try to write about issues like data quality data cleaning um model building because those are pretty generally relatable to people but if you write about this random small subset of thing in machine learning like a lot of people just won't know what that means or why it matters does that make sense you'd be surprised the subset is quite big i mean the the subset you're talking about i mean i for me the process is different from the from the way you described it's it's a bit more instantaneous it's worth more on the like i said on the spur uh so i i just sit for some time and then i just write in fact sometimes i'm not even in fact sometimes i don't even reread whatever i've written and just write a okay i'm just going to post that's good and yeah and i think it's i think what people figure i i think that message is important because you kind of have to figure out what works for you so right like it's a it's more of a matter like i know what works best for me and thinking about how to post but writings for the moment is okay too don't and people worry about how they phrase things did i use the right word did i decide on this perfectly i'm like if you worry about that it's going to stop you from posting yep just just post kind of learn about what happened when you posted and kind of think about things from there but i like i really think that's good it's probably a good example to see you and i'd post in different ways because there's no one right way to do this you kind of just have to figure out what is most meaningful for you and this was a fear for me as well like you mentioned right uh am i writing the right words not in terms of writing but in terms of making a video so i've recently started making videos on youtube uh which is which is uh something very new to me and uh i'm i'm very skeptical of how things will go and i'm like all the time so for me it was all the the the thing that was holding me back was is this video of good enough quality right like is it really looking good enough is this having the right things but then that was holding me back from even getting started so yeah yeah and now i've let go of that and i just put my curtains behind and then i just start talking so hopefully i'll get better at this yeah yeah hopefully but yeah but like yeah everybody improves over time like yeah you get better at it you have to practice though like you're not going to start out by being really good at writing and creating content you just kind of got to start somewhere sure yeah hopefully i can yeah improve on my video shooting editing skills and well uh i think i have one last question from my side which is what what is one advice that you would give to anyone starting out right now so starting out right now there's a lot of things going on in the world um but what's going to be important at the end of the day is again it comes back to this idea of look at look at your position as if you are an executive in a company and think about if you were to invest money imagine that it was your money that you were spending right as like a ceo what would make you say okay this person should be part of my company right it's like a and so that can come in multiple ways right in some cases this person just becomes really foundational and helps you understand and get your data in order right they help build the right foundation in other cases this person actually understands the business so well they help you actually think through what ideas to take and go forward with or sometimes they're just so talented in their specific area of data or machine learning and that it's easy to see the return on investment so if you think about that and it's and it's kind of uncomfortable and weird but if you say like okay why if i'm going to get paid this salary what justifies it right like what am i doing for the business to help them move forward and so i think it's useful in that way because to the business it's not about r or python or the tool that you're using typically it's about can you get to useful things and can you help us grow as a company right and it's a useful mindset um and it's really important amidst the time like we're in now where hiring is more limited but that kind of mindset and that approach can definitely make you stand out sure one last related question on that so you wrote a post on how you've read more than 5000 resumes so far and uh and so i think i'm not going to get into the post i'm just trying to ask you uh let's say you are reading yet another resume what are the immediate red flags you would see in a resume and how would somebody go about correcting them yeah red flags i see and these are and again i don't want people to take this as criticism because no one really teaches you how to write a good resume you just kind of are supposed to figure it out along the way but typically what happens when i see a resume is people treat it like a historical document where they list everything they've done at every previous position it's not that it shouldn't be bad um your resume is a marketing document you are the thing that is being marketed and so part of that means making sure to emphasize the things that matter most to the place that you're applying if you think about marketing if you just so i gave an example yesterday if you contrast two statements like i built this model and we used it versus i built this model in this amount of time and it saved this much money and it created this much opportunity for other people those are entirely different statements and i notice like so what stands out to me is when people can position things in terms of maybe impact maybe model performance maybe something that connects back to the business and you don't have to like i know it's tempting to write this will on a resume i have to write what all of my job description was you don't just pick the most relevant things um and so if you go so so if you look at it as a marketing document be like would i would i be interested would this peak my interest if i was a um if i was a hiring manager it can be a really good mindset to have but again it's kind of uncomfortable too every time i look at my resume i'm like if you don't know especially if you come from an engineering science background right it's not easy to to think of it from a marketing perspective but uh i think yeah i think it's very important there uh i think that's all for the questions that i have you