the inner workings of job search algorithms w/ subhash jaini

full transcript

auto-generated from the audio, so expect a few typos. the audio is the source of truth.

Alrighty, we are live. Let's see here, send this, send it now. I want to drop in the Discord, hold on to the Torque Discord and see here. Good morning, everybody. I forgot that I can share my LinkedIn too. Yeah, yeah, yeah. Although I don't really keep in touch with LinkedIn very often. Really? what's going on everybody good morning let's say everybody made nobody yet because we're still uh we're still trickling in you know people if you do tune in um one i buzzed all my hair off so that's the first um this is this is not normal um uh the second thing uh is you're probably taylor that is not your background yes uh listen i'm gonna put at t on blast here for a second Y'all got to get together.

I've had more internet issues at my house in the last two weeks than ever before. So I'm holed up at a Starbucks right now. So I did not want to cancel on Subhash today. I was super excited. So we are making this happen. This is not my decor here. I just want to preface that for folks. But it's been a minute since I've gone live. Subhash, you're, I think, my first guest in... I would say at least six weeks, I think, at least. So yeah, yeah. So I've been doing this show for six years and it kind of, it's morphed into kind of this like networking hour basically is what I call it with a guest.

But took some time off. It's been crazy busy. I've had a bunch of DMs like, Taylor, where are you at? It's like, listen, when you're trying to manage a team and cover as much ground as we are, you really don't have a ton of time to go live and stuff. But your boy's back. Subhash is today. I'm in Phoenix next week for a community meetup. We're doing two nights in Phoenix. And then the following week, I'm starting a job search series. Subash, heck, maybe bring you back on. But I'm trying to use a very beginner's guide to how to use AI in your job search.

I'll be using Claude. I talk to a bunch of people on a day-to-day basis. And basically, like... I was talking to a director of engineering yesterday. I was like, hey, man, he was in the hardware space. It wasn't quite SaaS specific. And basically, yeah, Ryan, he goes, I tuned back in and everything is different. Yeah, yeah. No hair. Different spot. So basically, I'm doing a quad in the job search, but I was talking to this director of engineering yesterday, and he's in the hardware space. And I was like, man, have you just thrown your resume into AI and be like, hey, what job boards do I need to be on?

And he was like, no, I haven't thought of that. And so like, I think there's some basic fundamental blocking and tackling that people can use AI for that isn't super rigorous, that can really kind of slingshot you on a really good path. So I'm going to start that series, it's gonna be an eight week series started in two weeks, you won't want to miss it. So, but again, as always, please say, hey, would love to hear from you. It's been a minute. I know we started a little bit early, so I'm sure people will tune in here in a little bit. But Subhash, just go ahead and dive in, man, because we have a lot to talk about.

You were already ready to go, man. You logged in backstage. You're like, I have so much to say. So if we want to do quick intros real fast, who are you? What do you do? And we'll go from there. So I'm Subhash. I'm a data scientist and I do random algorithms for random amounts of money. I mean, I love it. Nowadays, I just when things fail and they need somebody who can definitely guarantee that the AI works reliably, they bring me in. And I can be reliable. It can be reliable enough to depend on it. But it takes you actually have to understand the defense.

So, yeah, that's what we're going for. But like data engineer background, give me, give me, give, give the guests a little bit more context. Like how many years, where have you focused? Set the stage a little bit more. Yeah. Yes. Fifteen years. I've invented a bunch of algorithms. I'm usually the mad scientist behind the things. Um, I think my flagship product was for the organ transplantation system. And I had been there already a year, but it was really hard to improve what was already there. But I was able to get a big delta, but it had to do with filtering organs, but being transparent about what the transplant surgeons accept.

and then letting them do more complex filters. But really, it was just turning an algorithm into sentences and then telling the surgeon, you have accepted, you have rejected these a hundred percent of the time. Would you like to automatically reject it? And then have a conversation. And sometimes the filters were like over age fifty five and they were like, oh, yeah, our transplant center doesn't take anybody over fifty or like you got some other time. They're like, oh, no, that shouldn't be. And then it got more complex. That's crazy. That was the coolest thing. I've also done image recognition for construction. I've also done prevailing wage, collective bargaining agreements, recruiting algorithms.

reverse engineering, um, basically detecting like discrimination, a lot of lawsuits. Like if somebody is bringing out like a legal cases, they bring in me to just like, I'm a hammer. I have no strategy. I need a strategy. So they have to be like, Hey, the company we're suing needs to survive when we're done. Let's, you know, and I'm like, yeah, take whatever you want. Like, I'm just very open about it. Look, we could strike any of these angles. You coordinate it lawyer. You're the one who does that. Yeah, that's great, man. I love that. You're like a bouncy guy. You're like a bouncy guy. I love it.

I love it. Roxy, it's good to see you again. Charlotte, we're at capacity for Charlotte meetup tonight. Roxy's heading out to Charlotte to run that meetup. So shout out to Roxy. Hope you crush it tonight. Appreciate you, girl. All right. let's get into it so this introduction came from our friend stephen evans stephen evans is a recruiter in the richmond area i interviewed him a few weeks ago i think he was one of my last guests before i took a break if you have not looked at that episode i would highly recommend it um but so we made the introduction here so subash kind of give the context on what we're going to talk about today Oh, yeah.

So through the job application process, it was brutal. And so as a data scientist, I figured I'd do my magic against the actual process. And I found a ton of gotchas and things that people will not know. And so I want to let people know. And honestly, I try to do these one-offs, but it does take a lot of time and a lot of energy. And I'm hoping that Taylor can extract these questions and this knowledge out of me so I can pretty much refer this. So this is going to be like my dictionary. I'll point them to this YouTube link and then basically from the future just say, hey, this is my brain.

This is exact. And I'm telling everything I know. I love it. This is it. This is getting in Subhash's brain. I love it so much. All right. So Subhash, also, bro, I think you shared the workspace link. People are trying to join the live right now. So I don't know. I don't know if you want to take it down anywhere. I have so many people right now trying to join our live stream. I don't think I put it in the... Here, let me... Hold on. I've never had this happen before. Maybe it's in the Luma for some reason. What's going on here? I mean, it is...

I have so many people joining. Hold on. Here, I'm going to click this. And... Hold on. I'm going to join the event and see what it does. Oh, dear. Oh, yeah. Okay. So that is on our end. That is crazy. I did actually see the internal link. I did absolutely send the stream yards. That is crazy. I've never had this happen before. That was a misstep on our team. If you are watching this, go to our YouTube, YouTube.com backslash Torque.D-O-T-D-E-V or go to mine. But please stop trying to join the live stream. I'm only interviewing Subhash. Okay. All right. I will continue to kick everybody out.

This is crazy. Yep. There's another one. All right. So let's get into it. Subhash, where do we want to start? Let's just get into your brain. All right. So what do you want to know? There's a lot to it. In a job search process, there's a lot of things that people don't know. Let's start from the beginning. How should people approach it? What is an ATS? Let's start there. What's an ATS? How should people approach the ATS? Okay, cool. So basically, there's really two core algorithms. And if you were to have AI help you, there's... Well, actually, I guess there's three. There's the automation of actually applying.

There is the automation of your... score and then, and so like there's the automation of applying the automation of scoring and then the automation of actually finding the jobs. Those are three separate automations. Finding the jobs is challenging, but I'm going to mention that today. Um, the automation of actually auto applying, that's a little tricky, but I can, you know, that's, that's something that a lot of tools do. And then the, the matching algorithm, that's kind of complex and there's a lot of nuances there. So, um, Those three are the things that I would say are the three big categories where you would want to have different AIs.

Like, if you have a cloud bot, you would want three separate ones to do each of them because they're totally separate ideas. Okay. Okay. So, um, how, how do you think people should approach the job search right now? Cause it is a mess, right? You know, I'm out here saying kind of apply, probably not apply. Your network is everything. Try to get referrals. Like how should people be approaching the job search right now? Yeah. The network is everything. It's like, uh, a few years ago, I remember talking with a very, very smart advisor and he's like, so you're thinking the world is going to go to what you know rather than who you know and before with the world who you know it was a different world i think four years ago even seven years ago who you know huge even in my latest job search journey i got intros i got the i but then before my algorithm to match the matching algorithm um i got rejected even with the referral because they still have to put it through their ats a lot of the times And by the way, I'm going to number it.

I'm going to number like number one is going to be the job sourcing. Number two is going to be job Apple applying automation. Okay. Number three. So if I do numbers, that's what I'm symbolizing. And three. Okay, good. I love it. Matching score. Okay. So the matching score is important. So like the, it's important who, you know, cause there's some jobs that don't use ATSs, but. I think the big takeaway is that it's not just who you know. You actually have to score decently on these job applications or else you'll get rejected. Sometimes recruiters and staffing agencies can actually override that. But, you know, nowadays with Fusions, like Workday is now like, okay, before Workday, you were kind of like, you're a consulting, like a staffing company would come in and they're like a user.

But now with Workday, it's kind of like, well, we'll take all of it and you can add your greenhouse applications. And so what that's doing is that's putting everybody through the same scoring algorithm. And eventually it's going to be a situation where everybody will eventually go through the same scoring algorithm. But right now there's this mixture. So how do you write your resume for a scoring algorithm? Because even in my latest job search, Do you write the resume for what you've done? Or do you write the resume for what you want to do? Does that make sense? Like, I feel like the... It's a never answer.

I can't answer the philosophy. But I can add from a technical standpoint, what you need to do is, it doesn't matter what you want to do or what you've done. It's a matter of what they post. Because if you're looking at matching algorithms, it's about matching that job description. And like, I just opened up my resume real quick. And for example, I had Python in one and well, actually, I've already cleaned this one up. But like, Python, if you use the word Python, but Python frameworks is actually in and this is where the ATS really matters. the ATS is going to tell you what you should fill out because it doesn't matter what you feel.

What matters is what you score in that. That's crazy. I hate to say it. It doesn't matter. It's not a dream thing. It's not a situation of like, Oh, I want to be, let me make my resume look like that. You know, like, It's irrelevant. What's relevant is how close it matches. And I'll even say that some job descriptions actually put their benefits and their company description in the matching algorithm that they don't intend to. So some of the scoring algorithm has the word benefits or something. So if you want universal words, you want to have these because you can actually see in the code, you can actually see in the job description and then the benefits section is blank.

So like If we're talking like, I think ADP is that one. And again, like you can always just look in the source code or look at the patents to verify what I'm saying, but essentially different algorithms, different ATS algorithms are going to have different requirements, but you have to match it. It doesn't matter what you want or what the company wants. It matters what is matching, what is written. And sometimes in some exceptions, it's not what's written, But it's the scoring mechanism after, but it's heavily based on what's written. So, for example, a staffing agency will or an employer will or a job sourcer will post the job and then get an automation of what skills are required and they'll manually modify it.

And that's your scoring algorithm. If that happens, usually there's a mechanism in the back end of the JavaScript that if you get automatic rejections, it's there. If not, it's kind of a crapshoot, but you can kind of guess based on the job description should be pretty close to it because it's generally going to be from the text of the body of the job description. So Amelia says it means rephrasing the resume for wording of the job description. yeah but now wording it's not just the wording so like let's say for example this is ridiculous this makes me so angry by the way like this entire conversation because you should just be able to i mean it's ridiculous yeah it pissed me off so let's let's call out some of the obviously crappy things about a resume about a job description let's say for example it says it wants to um Let's say the whiteboard.

I forgot to erase my whiteboard. But yeah, it's just stuff, random stuff. Okay, so this specific job is looking for an engineering degree, okay? Well, Workday is looking, and Workday, now I'm calling Workday out, but Greenhouse has its own database. If you want to know all the jobs, sorry, let me jump. back to the wording so let's say they say python framework and that's the registered skill python is not enough it has to be python framework interesting it's actually and if you look you can actually validate it because if you look in workday like i'm calling workday out because its skills are pretty transparent but it's also the same thing others have um And I haven't touched this in a long time, by the way.

I haven't applied for a job since March. I just gave up. And I was like, screw this. And I'll tell you that story later. But that wording matters. And you can see in Workday's skill cloud, you can actually see after you fill out your resume, you fill out the sections of what you've done. You actually manually put in skills. Well, those really matter. Those are huge. And they do drive a lot of matching. But basically that adds to it. So it's not just matched the job description, but it's matched the keywords in the job description because there's another score in matching the job description because people have figured out that they can just jump, dump the job description in the resume.

And if it doesn't line up to a sectional connection like it has to either be in the experience tab or in the projects tab not in the education tab it has to be in the sport and sometimes skills but nowadays they're getting away from that because they want length of skills and they want level of expertise so there is level of expertise would mean either multiple jobs or a certain amount of late um and then so yeah so this is why like in the tort community like we do resume reviews and stuff we tell people even if it's a free even if if you're doing a project you're not getting paid for it it still needs to go into work experience because i mean this kind of validates that right is that it's got to go under work experience you either put it under under experience or you make a new section called projects projects will separate it and it makes the ATS combine.

But if you put it in work experience, sometimes it just pulls the latest and it does funky things to overlapping time. But if you put it in projects, generally projects is scraped in a way that it handles overlapping better. Okay. Interesting. That's a good note. David comes in here. Is there a difference between rewording and tailoring a resume? Or should we start using those synonymously at this point? Yeah, to me, they're the same words. Okay. Yeah. How would you define... So let me ask you. So I think... Yeah, I mean, David, I'd be curious to hear your thoughts because I would say they're the same thing.

I think if you're tailoring a resume, you're maybe making some slight edits. But if you're rewording it, it's more of a deeper dive. But David, I'm kind of curious if there's a delineation in your mind. Let me ask you this, Subhash. Could you hypothetically take a job description, take your resume, throw it in the chat GPT quad, whatever your daily driver is and ask it to basically make it ATS ready. Is that something that could help? No, no. So you're going to want to do a few things and this is for, this is for matching score, not for automating application, but the matching score. So what you're going to want to do is you're going to want to paste the keywords If you just paste the link, you can, from a technical standpoint, you can paste a URL, you can paste the whole job description.

But what you really wanna do is you want to paste the job and then you wanna say, pull out the keywords with like, a ninety-nine percent chance that they would show up. And then you get those isolated keywords And then you get those isolated keywords and that's what you're going to want to put. But again, it has to be written in sentences because it doesn't just show up in skills. It needs to show up in linguistic sentences inside the actual experience. So you have your work experience and then you have the company you work for. That timeframe you work for, if you put a keyword like Python, And Python might need, there might, and to give more nuance to it, Python may need to appear three times, like depending on what they call.

Right. Yes. There may be a count, but really if it has the word Python and you worked at that company for eleven years, it's going to mark you with eleven years of experience of Python. So that's kind of how dumb it is, but also how smart it is. You can't just put it in skills nowadays. can, but if you want a better score, you're going to want that timeframe because oftentimes people ask for a certain time range of that. So this kind of continues to bolster, because again, my brain is running a thousand miles a minute while you're talking and kind of how we're educating folks in the tort community about this.

One thing we say is that your resume needs to be littered with key words, not only in the actual bullet points, no, not only in the technical key words, but also in the bullet points. Now, I'm still pulling from old school recruiting days where you just control F stuff. In my experience, I've seen folks who are a Java developer but only have Spring Boot in their bullet points and not Java. And I've told people, a human's probably going to reject you, a dumb recruiter's probably going to reject you because they don't know that Spring Boot's in the Java ecosystem. And it sounds like the ATS is kind of the same now.

It's pretty much they have some crosswalking, but most of them, now the ones that are AI driven also have the same holes of AI. So for example, and it's because there's a workflow that is, and I've had this discussion with somebody who was like, AI isn't used in any of the job, any credible situation. And it is. it is and so like i will say there is ai like chat gpt ai and then there's the stuff that i've been working on for fourteen years i've been for twelve years i guess whatever um since twenty fourteen which is why i always mix up twelve and fourteen yeah yeah yeah uh but but that stuff is definitely used in that in fact there's a richmond company i forgot what it was but it was Data.

It's one of the back ends. Daxstra. Daxstra. That's based in Richmond. Woo. So but there's other platforms. There's text. Oh, man, I forgot. I had my code. Text kernel. Text kernel. Yeah. Yep. Text kernel. Thank you. And so like people are using it. They're just using different wording, different AI. But but yeah. Really, you have to have the skills that are niche and specific. You want to be putting those in there if they show up. So, you know, if I'm a job seeker and I'm listening to this, I'm like, this is overwhelming. Yeah. How do you, like, based off what I'm hearing you say, do you think that we should be editing every single resume that we submit?

It depends how much you want the job. So if it's a specialty job, then I would use my algorithm and I have a code based algorithm, but it basically workflows out optimizing, but it does it section by section. And so that's why I'm using code because I want it every section to match up and then it to have a score and then it rolls up to the work. Then projects get certain amount of points and then it all kinds of, and I don't want a hundred percent. If I get a hundred percent, it's no good. So I need to drop things off. That's crazy. That's crazy.

Oh yeah. Yeah. It definitely works out that way. So, but what you can do, and this goes to step one is job sourcing. If you job source a lot of jobs, you take those job descriptions, then you pull out all those keywords, then you kind of have a resume that works. Then you can use a traditional automation resume thing that will upload and it'll get you really far too. Okay. Okay. But the other one scores you really high. The, the, when you optimize for resume matching, that's when like, if somebody makes an intro or you really want to work at a company, then spend the energy.

But otherwise I would actually just work on the bulk and getting there. Yeah. Yeah, I mean, Seth, the games that have to be played is wild. Yeah, I mean, listen, I don't like it. I don't like it. I still think this goes to making contacts beforehand. I mean, what are your thoughts on that, Subhash? It's irrelevant. I really do think so. You really think so? I really think so. And I've permitted for about three years, so I don't really go out and talk to people. That's fair. I am very networked, and it didn't help at all. And cause at the end of the day, it all had to go through scores.

There's testing too. I suck at test taking. I scored like, twenty percent in Python from a very close friend. Like she's, I would even say, I call her my best friend. And so like, but we have a group, there's three entrepreneurs that we just were best friends and she tried to get a job for me and resume matched and like, and like the resume was okay. I didn't have my script back then, but she let me take the actual test. And then when I took the test, I failed. And she's like, I've seen you fix my problems in Python live. I'm like, I can code, I can do the real thing, but now I need, there's the written test and verbal.

And then there's the interviews where they're trying to get gotcha and they're just trying to filter. So I am blessed enough to have been done enough projects in my world to get projects and make this insurance model of a business. But otherwise it's very stressful to be bound by these ATS. So how do you know, is it accurate that every job you apply to has an ATS? Is that, is that, uh, no. Okay. Talk about that. It's basically the ones that I optimize for and I'm trying to find my code. Um, there was greenhouse. I would focus on greenhouse lever. I'm trying to ATS. Oh, I opened it with Notepad++.

That's why. Let's see. Workday and Workday has Workday one, three, and five are the ones I would optimize for. I would separate them. There's Greenhouse, BHR, breezy HR, LinkedIn, use it. LinkedIn's kind of a pseudo ATS because it integrates with other ATSs, but then there's a workable tool, but it's, it's, you know, they're, they have their own thing. Um, then there's, I put as none, like technically there's UKG, but it's actually very basic word matching. Um, and ICS, which is, um, the old one. There's an old one that appears. Yeah, I forgot what it's called. Let's see. So when you're reading out these ATSs, I see IMF, yeah, and Tulane.

How should people approach their job search, right? So they listen to this live stream. They're taking notes. I'll send it out in an email newsletter soon. You mentioned these ATSs. How can people better themselves? Like what would be some actionable pieces of advice for folks? Source real jobs, not ones like Glassdoor and all that crap. Real actual company jobs. From the company jobs, yep. Again, and actually I have them. These are the algorithms that are the back ends that I've separated. It's Workday, LinkedIn, ICIMS, Greenhouse, Lever, Breezy HR, Jazz HR, which actually just used Eightfold, Taleo, Text Kernel, None, and Daxstra. Those are the algorithms.

The specific systems that post jobs are going to be Greenhouse, Breezy HR, Workday, and Lever. And, oh, Ashby HQ. Just literally lean on those and don't do any more. And then you find real jobs. then you do that whole keywords finding. And if you lack the skills, that's something you need to know. Because when you do your searches, if it has that keyword, you can automatically source jobs. And if it has that keyword and you lack it, don't even apply. Just move on. So that's one thing that's good for you to know. And then it also helps you like job description optimization. It can help you when you look at an aggregate, you can find the right keywords.

Okay. Okay. Awesome. Awesome. What else, man? I have some questions I prep, but I'm trying to get as much out of you as I can. Is there anything else pressing that we're missing right now from an ATS job search side things? Yeah. Yeah. So the key is a lot of times there is search functionality that will Google jobs and most job boards are going to be in your way. You want to push those out of the way and you want to go to the company ATS systems. And I've been talking about that consistently too. Okay. Yeah. Because from an algorithm standpoint, I'm watching the world.

So I like to watch everything, but I'm doing it for my own purposes. And at this point, I'm not here to help people find jobs. I actually made a little website one time. Honestly, I told you about this, but like I made a website because I was like, oh, cool. I can let people use my tool for five hundred bucks. People had so many questions. And I mean, I feel I understand the stress of being unemployed. I can't get that client base like I can't guarantee. And so like I can guarantee certain people find jobs. And once I look at them and see their profile, if they're like me, but they just need, it's usually the sourcing is what people need or the matching score.

And then I can let them use my tool and my tools for my costs. It's eleven cents an application. Eleven cents each time I make that resume. Um, and so it's not worth it if you're just, if you've been, it's not worth it for everybody. Um, but yeah, that's the key is it's either in the matching score or the job sourcing. Usually the automation to apply is honestly something that I, if I like went back and I was serious about finding a job, I would outsource that and say, Hey, let me hire somebody to just upload resumes for me, but they'd have to run it through this custom GPT or this quad bot that has this Yeah, yeah.

What else? What else? What else have we not covered yet? Yeah, you can also have trap words. So inside people's resume, they're putting in inside jobs description, and I wish people did it more, but they'll put a line that says like, something about automation or whatever, and they'll put it in there to kind of throw automated, automatic, um, resume generators. And it will, it will fool everybody. Like it will fool most people. But, and that's the thing is you would just have to see how they're worded. I'm not going to disclose them. Cause I think it does go into a dicey territory, but you will, especially when you start, like, if you use automation, you will start to notice certain keywords or certain bullet points.

Um, yeah. That's crazy. So, and then why do companies put those trap words or phrases in the job description? Because they're dealing with the whole problem of too many resumes coming in. And so like I've helped somebody filter out in a different way, but basically I told them, I don't want to filter your filtered resume. I mean, I did that a long time ago in like, but like this is with modern stuff. I'm like, let me make this algorithm to filter, but let's bring in all the resumes or all the resumes that at least had these three or four things. And even something like asking for an engineering degree and thinking you're talking about computer science, engineering, data science.

Well, data science isn't a marked degree or an official degree. So greenhouse has no way for somebody to fill it out. And so for example, there's these quirky things that in the matching algorithm, you kind of have to change your degree name. And that's another thing that is tough because Are you lying? But there's no way to submit a resume if you do. And also the job description, it's clearly a data engineering role. Why would they mention engineering? Well, you know, like it's, it's like, you know, it's wrong. So you, it's tricky, you know, like, yeah, you got to play the game. You got to play the game, get, get in.

And then scoring mechanism is important. It says, how about we list that we don't have the skill? Will the ATS understand a negative modifier since no one would hypothetically do that in a resume? It will work. It will work. If you mentioned you don't have the skill or you tried to use, it's true. I tried to use Python in this, or I started using baby Python scripts. it'll actually show up as python in that thing like it is it is that dumb wow that's crazy that's crazy all right i'm gonna get into some a little more questions here um uh let's see here we've we we've covered a lot is there any place people can go to to to maybe see how their resume scores Yeah, you can sign up for text kernel.

I think actually lets you trial a text kernel, but they have two versions. They have their traditional and then they have their AI based one. Realize that both of them are linguistics AI based like any resume parser. has algorithms to not only read it in, but to parse it and to do it. But one is using like chat GPT AI and the other is not the chat GPT AI is actually a better, but it also has holes in it. Like it has some big holes. Okay. Okay. So, so, so you can go to text kernels website. Yeah. I think like kernel will actually let you have a trial, but yeah, I would just sign up for trial for, um, um, yeah, I think it's text kernel.

Let me see. I remember. Oh, no, it wasn't text-frontal. ATS parser trial. I forgot which one it was. Maybe it was deal? Deal. Deal. D-E-E-L. Workable doesn't really have a scoring mechanism. Okay. Job scan has a little bit of it, but I'm trying to think of the one. I thought, I remember text kernel. I feel like I did a text kernel trial, but try to find an ATS system that, so you can just see what I'm saying. Like you'll notice that if you put the word Python in one job and put the word Django and expert developer, but don't use the word Python, all of a sudden you have a gap and you're not continuous Python.

Like you'll see all the quirkiness that it puts you in. I think that's great. Yeah, everybody should try that out if you're on the job search for sure. Let me ask you this. We've obviously talked about optimizing resumes for keywords. What do you think the common culprit is for getting rejected? Like, based off your experience, is it literally just the keywords? Is it, I mean, like, is there any themes that you've seen with your friends or yourself? It's a score thing. I mean, really, it's just, I mean, that's the reason people get rejected. It's even the weird cases where somebody puts a trap. I don't see those traps that often, honestly.

I see them like, they're kind of rare, but yeah. Based off your experience with ATSs, do they match from an algorithm perspective relatively similarly or are they pretty off base? Like if you can get it close to a... They're very different. Oh, cool. Cool. Yeah. And how they count like ICIMS is old school. So it actually doesn't really count time and skills as much. but it will count keywords and it does count. And so like LinkedIn is more keyword based because they have to deal with lots of blobs. They don't have like the functionality. Greenhouse actually has different sections, but then they let you, but they give, they basically put it on the other things.

There's something called Phenome, which is a tool that people stack on top of Workday and the like. And that has a score, like, really complex scoring, which it does pretty good, but it's also exposed in the source code. A lot of times you can see people's scoring mechanisms by viewing source of the job description. A lot of times. No way. No way. And sometimes you might need to actually be in the, like – I forgot what JTS you click when you click apply, then you view source because then you can actually see, and I've actually seen contact information of the recruiter of the, of the, I've seen admins of the account.

I know what work, what plan they're in. I've seen how many people have applied, how many people have viewed it from indeed. Like it's actually, and then, Oh, by the way, indeed, also it posts in the source code. You can actually, it's not in the source code. It's, You can use AI to help you, but it's basically in the DOM, it has a data object that holds the scoring mechanism. Because in Indeed, it shows how much of a match you are. So you have to use console.log, but you can actually see how it's scoring you. And then it's usually tied to ICI. A lot of people who use the older ICIMS and...

Taleo, they are using UKG. That is Indeed. Indeed is a lot of those older jobs and a lot of work. You filter out the workday and then you've got these three main old school ATSs. And Indeed is essentially their filter. Indeed is their scoring mechanism. So even sometimes you'll apply on Taleo or on the original company one. And Indeed is unique in the sense that people will port it over. So it's both an ATS, but also just the scoring mechanism too. So, and I don't, and at that point I'm like, what do we call it? ATS? What do we call it? You know, like it gets, that is crazy.

It actually sits on top of other tools. So F or P H N O M. And that's the one that exposed a lot in the source code. I've never interviewed somebody on this topic. This is crazy. This is mind-blowing. I've never met anybody that's done this, that nuked it. And I didn't want to share this at first because I'm like, hey. I remember I even shared it. I used to share posts like this. And I got a call from a really good friend of mine saying, Subash, you can't post this shit. You're making anybody who interviews you think you hacked in there to get the interview.

And I'm like... Oh, I forgot about this. So when I was really looking for a job, I stopped it, but then I'm just annoyed at this point. So it's been long enough. I've got my money. It's straight. Um, Yeah, so at this point, I'm like, screw it. But PHENOM, in the event that they're using Workday or Greenhouse, but their actual company job posting is on their website. So in the sense that they have the tenant that you'll see on their website, they typically put the PHENOM scoring mechanism on that page, not on the Workday because it doesn't exist on the Workday.

But if you look on the company site for that same job, you can see the score and you'll see like, you'll see like if they say five to seven years of python don't you dare put ten years yeah and you'll see how much they take away you can really oh yeah they can see seniority level sometimes you'll see the hiring manager's email um so yeah like crazy crazy what you can see so let's let's kind of let's as we kind of slowly start to wrap up here guys if you have any questions please ask them uh also please go connect with subash on linkedin uh

obviously he's a wealth of knowledge um you know you've talked about you've had problems finding a job which i think is hilarious because i think people who watch this are like how does this guy not find a job yeah yeah so talk to me talk to me about how you pivoted to your own thing because i think right now like i think the market's right for people with ai and spinning stuff up and you can create a landing page you can get ideas out can you talk just about general career stuff for me oh yeah so the reason i would love to have a career

i would love to be just a regular employee at a job but when i got this algorithm working i got nineteen interviews one week and i was still running it and i got sixteen the other week and then twelve and by the way this was goes from once a month to like it was chaos but then i got rejected by like thirty people and that sucked so yeah yeah the job market is it is ripe if you're using these algorithms and these techniques There are people available.

They will talk to you. But as far as the if you're not, you're missing out. It's definitely true that you're missing out if you're not optimizing your resume for the submissions. What are your thoughts on just AI in general? I've been talking about AI in your career now since I gave a keynote in October of last year. I've said the gap is going to close. I mean, could you say the gap is officially closed where if you're not at least using AI in some sort, that's a red flag now? Do you think we're there yet? I don't know, actually. I have no idea about that kind of thing.

Anytime we talk philosophical, but I do sweep up things. And so like, I know people who are early adapters and are thought leaders in AI and they're producing shit that I have to clean up. And so that's a constant thing where like- That's hilarious. Yeah. the more data that comes in, it'll get better. I'm like, it won't, you're using an API call. It will not get better for your, and then even when they custom train it, they're not doing like, they don't know what an ID is or rejects or cleanup scripts. So they're not, so they're just like shoving data in or rag they're doing search and they're just shoving every file in.

But, um, I think, I don't think we're done with those days. I think those guys are still going to be on the top list of everybody's AI thought for a few more years. I, I, I, I give it anywhere from one to three more years of those people. But eventually, rubber meets the road and people don't get results. People don't get the rankings. They got rankings for the first month and then nothing more. Or the automation worked and then it broke this one specific way and it was unable to be better. And that's when people are really struggling. And they're like, oh, we need somebody who really – understands what powers we're using or when they depend on their business on it.

That's like, yeah, I don't know. So that's from a tactical standpoint, but I imagine there's going to be a lot more, you know, snake oil sales. I think there's a lot to be done. I think, and like, and not to say that's always bad, but, Because even if people are selling snake oil, because there's some truth to it, people are like, well, maybe I do need oil for this. So they're finding some use in the oil. And then somebody real can come in and do the real heavy lifting. And the organization can still improve, but it could also harm if they're not careful. That makes sense.

If you can't code, AI is a very dangerous game to be playing in. Yeah, yeah. Let me ask you this. Obviously, you are doing your own thing now. Any best pieces of advice for people who want to start kind of their own business? Just general do's, don'ts. It sucks. I hate the piece they're fanning. I hate it. I think going with the insurance model has been really helpful for me. But the only reason I can do it is because I have such a wealth of experience that I can go in. Sure. any problem like i'm sure like sending out birthday cards fixing an hr thing defend a lawsuit like all these things and like yeah i don't know what to do like business-wise so okay well there you go there you go i love it i love the honesty i love the honesty all right so um as we wrap up here any last pieces of advice for folks Yeah, I think that matching code, just to confirm the path, is you get the keywords, you find all the keywords, you get that, and then you put it in a very simple file for AI to understand.

And also, if you can put your resume in something called a YAML structure, Y-A-M-L. And you put your resume in a YAML structure and then you give that to AI. So AI is not dealing with reading a doc X, reading a PDF, parsing it, none of that. It understands at the core who you are and what you've done, what timeframes. And then what you do is when it generates a resume or generates sections or answers questions, it understands the deep ends of who you are and what you do. right so to add on this and i'm kind of thinking about the youtube transcript to come up with the newsletter so i'm thinking ai is going to consolidate this so again we are hyper targeting company job postings not job board job postings yep we are creating files around what again You want to have your core resume.

You don't give AI your resume. You give them a YAML file that has the information that your resume has, but it's cleaner. It has no formatting weirdness. Great. And then... what do you do to cross reference ats do you have like skill md or do you have md file like what do you what do you have for those and that's where at this point i'm still calling that like kind of like my proprietary science but it's okay i've just looked you can google search the patents and they're a little bit hard to read but with ai's help you can possibly do it but the patents and the actual ATS you can see.

And the way you kind of see how you're doing is the instant rejections or if they reject you in exactly one day, you know that that's a rejection. So you can make your own ATS. And also you may feel like, oh, I looked on a Reddit thread. If you apply at Workday, this does better or that does better. And and then also and because we've been talking about a lot of the matching score what you want to do is when you up workday with a job that you're not applying to but you just if you want to test it out you can apply to a random workday job and you can see how it parses your name your first name your last name your address and you might need a reword or reformat your resume to optimize for that and then and then see if it works on greenhouse and see if it works on lever and because it because yeah that's what that's that's it's a whole thing right So let's branch out that technical side for a second.

So very basic company website, actual jobs, edit your resume to match the specific keywords it's using and even specific timeframes of experience. So if it says five to seven years of Python, are you specifically saying five to seven or could you say seven years of Python? I actually, in mine, I will actually try to make it in that range. And that gets complicated, it has a chance of failure, because then you need to have not just a per job, then you need to have an aggregate algorithm. And essentially, that's that makes the scoring complicated. But if five to seven, even if you have ten years, if there is a punishment, it's going to be small.

And It's easier. It's not, it's over. It could be over engineering potentially. If you're limited on your time, just screw it. Just put all the Python that you can. Okay. Okay, cool. But yeah, so. So basically we're kind of, if you want to dumb down the conversation that we've had for the last forty five minutes, it's very similar to what we've been saying to our community. Edit your resume to match the words on the job description. If you're going to apply, apply directly on the company page and make sure you tailor and or reword whatever David said specifically to that. And it's going to take time, but that's what separates you in a nutshell.

Yep. I love it. I love it. Awesome. Track. And then also on the matching algorithm, track every submission you do. And there's, you score something like you could actually make a Python script that, or a Google script that basically when you get a job from an application, it goes in and stores it on a Google sheet. Like you can make these scripts that make your experience much easier to track. Yeah, absolutely. Absolutely. All right, man. Well, I ask questions at the very end. We're a little rapid fire here. So let me pull them up for you. I haven't done it in a bit, so I got to find them.

Give me a second. Where are they? Should have had this prepared. Let me see. All right. For some reason, the note's gone. So I'm going to try to go from memory. First thing, what is one tool or technology you can't live without right now? Python. Python. Love it. Love it. When you were younger, when you were in school, what did you originally want to be when you grew up, have grown up? Restaurant. Hands down. I still want to run my own restaurant, and that's why I want to build my own thing. But yeah, maybe a food truck. I love it. Any specific cuisine. Yeah. Pellets.

Basically a random amount of filling with some sort of wrap of some sort. And so like, and then I'm working. Sounds great to me. Optimize nutritious, optimize food science. So it cooks in the middle and then you inject it. So that way you can avoid cross-contamination, but also be able to pre-make a bunch of pellets. I love that. That sounds great to me. Sign me up. Next question. If you could go back in time and give young Subash any career advice, what would it be? Uh, probably not start a business. I think it black, it gave me a black eye. And so like, I think, I think that's, what would be the career view, but from a capability standpoint, my.

my like skill set is insane. I wouldn't have been able to skill up as much as I have had I not had my own business. So it's hard to see where I'm at. And so now that money's okay, like, I don't know. I don't know. I think things have been great. I love it. I love it. Sweet. All right, y'all. Thanks again for hanging out with us today. For everybody who tried to join backstage, I apologize for that. That was on our end. That was crazy. Feel free to share this. I will say this. In the six years I've been doing this live stream, i've never done an episode like this so if this resonated with you definitely share it go connect with subash um but again thank you all for hanging out subash stay on backstage so i can say bye to you um and uh y'all have a good one and we'll see y'all in two weeks for the claude and ai job search series peace