want with it. All right. So, as mentioned, this is a live build, meaning I haven't actually done the building yet. All I've done here is laid out a rough scope of what I think the system is going to look like. Then eventually, we'll dive into NAND and take this totally blank canvas and turn it into a workable system. So, just want you guys to know that I do this because I consider it more informative. I think that leaving in detours and bugs and stuff like that, it's a little bit more honest. You guys get to see like what the actual process looks like. And yeah, you know, back when I was learning how to build things in NAND and other no code tools, these are the sorts of videos that I always look for. So, I figure now that I'm at this point, I might as well make these sorts of videos for others, too. Okay, so here's how I think that this is going to work. We're going to start by getting some sort of resume into a templatable format. What do I mean by this? I just mean something that for the most part is just text, something that, you know, might have some light formatting, H2s, H3s, that sort of thing. but for the most part something that we can feed into AI and have AI reproduce for us without having to like add a bunch of complicated steps in between. So what I've done here is I've created a templated resume format for myself. And the only kind of like interesting pieces here are heading 4, heading 3es, and then I think that's it. Just some bolding and some metallics. And I just had a generate this. This is not like actually accurate whatsoever, but I want you guys to know that you guys can actually create whatever sort of ré template you want. This is just what I'm going with because it's simple and easy for me. Okay. So, after that, what we need to do is we need to scrape job postings using LinkedIn. Now, there are variety of tools that are available to allow us to do this. The one that I usually use on this channel, it's called Appify. This is basically just a marketplace setup where you have a bunch of these different people creating scrapers, uh, publishing them on here, and then charging people money for it. And because it's a marketplace, that means that they tend to compete, which allows us to drive prices down. The reason why I'm not building my own scraper in this case, while I think you can do basically anything in 2025 with technology, and I could most certainly build a scraper if I had enough time and energy, it would take me a lot of time and energy. Userenerated content platforms like LinkedIn, Facebook, and stuff like that, these have basically the most sophisticated anti-scraping protections on planet Earth. So rather than mcking around with having to, you know, solve that problem, I'm just going to abstract it away and use something that somebody else has developed that has the time and the energy and is incentivized to do it and then I'm just going to pay them a little bit of money for it. So I see one right over here called LinkedIn job scraper PPR. I've used this one before. So that's what I'm deciding to use for this as well. Developed by Curious Coder. It's a dollar per thousand results. Pretty simple and easy. And then from there, what I want to do is compare the incoming job against a bank of skills and have AI filter. The idea here is like obviously not everybody is going to be capable enough to apply to every job that comes their way. So we just want a simple but flexible way to filter out jobs that may not necessarily be in our wheelhouse. You know let's say I'm an engineer and I don't know for whatever reason my search is crappy and I get a job for house cleaning. Like I want a way to filter that out. I don't actually have to apply to said job. Right? Once we're done with that filter step we'll customize the resume to match the specs. Then you know the question is how do we actually do the application? A couple different ways to do this. So, you can totally build out some sort of like browser automation to go out and then apply to the job on, I don't know, Indeed or LinkedIn or whatever. This would again take a fair amount of time and energy. And while I have experimented with approaches like this and I made them work, I don't really want to go through the rigomearroll showing you guys what that looks like in a video, rest assured, it takes a lot longer than I probably can fit realistically into a good YouTube clip. So, what I'm going to do instead is I'm going to grab the website details of the company that is making that job post. I'm then going to look for contact details of decision makers. And once I have a contact detail of a decision maker, I'm going to send them an email using a short punchy pitch. I'm going to attach my resume in there. And I'm going to say basically, hey, I'm the person for the job. Okay? So, I can do the thing they're asking me to do. And because I'm going to be applying for technology jobs, I'm actually going to try and show them like proof in the pudding. So, because my system is an automated AI based system that, you know, does all the job applications, my hope is I could say something along the lines of, "Hey, I know you're looking for like somebody proficient in AI, well, I actually used AI to like build out a pipeline to allow me to apply to jobs just like this one. " So, I figured I'd rather show you than tell you. Happy to run you through what that system looks like. This is a little bit more, I don't know, if you want to say daring. It's certainly not like a corporate job style approach, but then again, I think if you want uncommon results, you have to take uncommon approaches. Okay, so that's what we're going to do. That's what I'm going to try to do anyway. We'll see how well we can stick with this. First thing we're going to do is we're going to go over this ré template very briefly. And I just need to find a way to connect this to NAN. So I'm going to go back over to my NAD workflow. I'm just going to call this like AI automated resume. Just call that for now. Maybe I'll add a tag and I'll say NAN course. And then over here, what I want to do is I just want to see if I can grab the Google Doc content. Okay, so I'm going to get a document here. In order to connect to this, what you have to do is you have to create a Google Cloud Console account, which can be kind of a pain if you haven't done this before. Rest assured, there is ample documentation out there on how to do this. Basically, you start by creating a Google Cloud account, and then you just follow the steps. You enable the APIs, configure your OOTH consent screen, create your Google Oath client credentials, and then finish it. At the end of it, you're going to get a client ID and then a client secret. These are just what you paste in here. Keep the OOTH redirect URL the same, and then sign in. Okay, so once you're done with this, um, what you do now is we're just going to get the document. But where is the document? Well, we need to feed in the document ID. And for future reference, anytime you see a long weird ugly looking string in a URL like this, odds are that is the ID of the thing. So in our case, it is slashdocument slashd and then an ID. Odds are this is the document ID. And I find if you just throw stuff at the wall, a lot of the time it sticks. So, we are executing this run and voila, we now get the content. So, we actually get the content of this whole document, which is great. It's exactly what we need in order to feed into AI. And it looks like we got all of it. Yeah, we did. One thing I'm noticing we're not getting here is we're not really getting the formatting. So, I'm not seeing like H2, H3, H4, whatever I had before. That's not the end of the world. Let me just click simplify and see if we get more data because this is a little bit too complicated for me because it includes too much formatting. So, I think what I'm going to do actually is I'm just going to roll with the simplified version of this, the one that just has like a content variable, and I'm just going to have AI like rebuild the format every time. It's not ideal because it's not going to be the exact same format, but my hope is, you know, it's going to be able to interpret or infer what is the heading here to hiring manager, obviously. here? Professional summary, obviously. So, I'm thinking that's probably going to be sufficient. I don't know if it's going to work, but we'll try it. Okay, cool. So, now that we have the data inside of NAN, next question is, how do we scrape job postings and then compare? you know, maybe the resume against the job postings and do something. So, I talked about using Appify before. I got a scraper right over here. I'm just going to click try for free. I've already created an account and you guys can actually get 30% off using the code Nick 30 below. So, feel free to use that. Uh, I don't get any money from that or anything. They've just been gracious enough to give me a code that uh makes it cheaper for people to sign up. What we do here is we add in a LinkedIn job search URL. So, the question is, what the hell is this? Well, we need to go over to this service, LinkedIn jobs. And then what I'm going to do is I'm now just going to like type in what I want that's interesting to me. The reason why all of these Zapier jobs and AI actor jobs and stuff like that are popping up for me is just because, you know, like LinkedIn knows that I'm into AI, so it's just automatically assembled this for me. I don't really want anything local. Like, does it have to be in Calgary? No. Um, can I just go United States? I just want like USwide searches, you know, probably remote. Title, skill, or company? Let's just do a automation over here. Now we're talking. These are cool and not all these companies are huge, right? Like I think there's a much higher likelihood of you actually getting a gig at like one of these small to midsize businesses than I don't know Zap year or something. So this is really positive for us. Okay, cool. So from here we can add some additional filters. So anytime past month, past week. Why don't we just do past month? I mean, you know, like a month is quite a long time. So maybe actually we do a week. Okay, looks like we have 800 job posts in the last week. That's some scale. I'm not going to filter experience level. Should I filter salary? Go 100,000. How many? 64 out of the 700. Wow. You know, I think in reality, most of these just don't show a salary at all. That's why barely any of them are popping up. So, let's leave that blank. Company, no. Remote. This is interesting. If I click remote, how many? 18. Hybrid. 104. Okay. So, the vast majority of these are on-site, obviously. Um, so if you really wanted to get like specific jobs in your area, you would probably have to, just to be clear, attach a search specific to an area, not necessarily just what I'm doing here. I'm going to proceed with this just cuz I want enough data to actually run this. But yeah, just letting you guys know that if you guys go remote, you only have 18 results as opposed to the 795. A lot of companies are still willing to work with you on a remote basis, even if you know, they don't necessarily say so right off the bat. But anyway, let's just give it a go. What am I going to do now? Well, I need to test and make sure that this works. I think a lot of people sort of put the cart before the horse. They put the forest for the trees or whatever and then they just like try diving into NAN and making it work. I always just like verifying that things work on the platform that I'm using. So, in this case, I'm using Appify. Obviously, what I'm going to do is I'm just going to scrape company details and then how many jobs needed? Let's just do 10. I'm going to click save and start and let's just actually find out what happens because who knows, you know, sometimes there are bugs directly on the dashboard. if there's a bug directly in the dashboard. I don't want to have to be thinking that it might be an NAN bug. Okay, so you can't use this actor for scraping less than 100 records because it's not efficient. See that? That's a bug right there. Right? Like I wouldn't have known that would have been a problem had I not put this in. Looks like we are doing some scraping now. Although it is taking its sweet ass time. Now I should note that I fed in less than 100 jobs and it looks like it's now outputed some which is cool. What are we getting here? Okay, so first of all, benefits, actively hiring. That's obviously a benefit if you are looking to apply to a job. Company description, that's good. Company employee count. This is great information. Company LinkedIn URL, which is nice. Why are we getting tons of UK jobs? The first two are UK. Company logo. We got the company name. website, which is great for us. Looks like all of these have websites, which is awesome. If we didn't have websites for this, that' make our life a lot harder. Okay. And yeah, this looks pretty solid to me. Looks like it's going a little bit above and beyond scraping even more jobs than I thought were in the list. There is usually going to be some sort of discrepancy between what the front end here shows you and then what an Appify actor scrapes sort of behind the scenes. And that's just because they're not using like a cookie for this. They're not actually signed into my account. So, their screen is different. It's, you know, it has some stuff in there that maybe get filtered out before it hits me. But that's fine. Okay, great. So, we've now scraped all of these jobs. This looks pretty good to me. We have all the data that I was personally looking for. The main one is just the website cuz from the website, you can get everything else. What I'm going to do now is I'm just going to find a way to take this actor run that I just did on Ampify and then move it over to NAD. So, let's just go over here. I'm going to type an HTTP request. And uh I've done this quite a bit in the last few videos, but just, you know, it bears repeating for posterity. Anytime I don't know how to deal with the docs, I just type in the name of the platform and then API docs. From here, my main goal is always initially to look for authentication guidelines. So, Appify is a really good API. So, if we scroll down, authentication is one of the very first things we see. So, here it tells us how to grab our API token. We go to the integrations page in the Appify console. Then, it even tells us the format that we need, which is authorization bearer format. There's even a little section that teaches you how to use authorization bearer format. Not all APIs are this good. You know, documentation is going to vary depending on what you're using, but Apify is pretty solid, which is one of the reasons I use it. So, I'm going to head over here. I'm going to create a new token. And let's just call this AI automated resume. I'm going to create. What we have now is we have the API token, which is awesome. So, if I paste this in, this is my actual API token. Should probably delete my old ones, huh? Can't tell you how many times my old ones have gotten run up by people that just like watch these videos and go, "Oh, did Nick forget? " And turns out Nick forgot. Anyway, so how do we actually convert this into an API call with the token that we have? Well, I've used this quite a bit. So, I know which endpoint to use, which is logically like let's just work through it in our heads, right? And even if I didn't know, what am I trying to do? I'm trying to run an actor. a scraper, which they call an actor. So, what am I going to do? I'm going to look for the end points here that say run actor. And I see one here called run actor. Odds are that's what I want. You know, there's run actor synchronously with input and return output. get data set items. run acts synchronously without Odds are I'm going to want some sort of input. data. So, you know, even from first principles here, one of these is going to be one of the endpoints that I would probably use even if this is the first time I was using the API. To make a long story short, there's no way to know for sure without actually just going ahead and using it. So, you know, like we just have to get the authentication up and running and then we just start using endpoints and we just see does it give me the information I want. I'm going to kind of skip ahead a little bit because I know that this one does. Once you find the endpoint, um, how do you actually convert this into an NAND thing? Top right hand corner, you just click this little button. Okay? And then, well, actually, there's one more thing I want to do. You guys see up here, right over here where it says uh actors/ whatever/input, this is the ID of the actor. So, I'm going to go back to my ad docs. I'm going to paste this in because this is required. And then bearer token over here. I'm going to go and I'm going to copy this in a automated resume system. I'll paste that in. Okay. Then what do I do now? I copy this. This is the curl. You always want to look for the curl or the rest API or the direct API call. And you just want to paste it directly into the HTTP request as follows. When you do this, Naden will take all that data. It'll parse it out and then it'll update all the fields for you so that the HTTP request automatically works. This is one of my favorite features of NAD. It's one of the reasons why I think that they've blown up so quickly. They just make um actual API requests pretty easy. It looks like my token didn't copy over for whatever reason. So, I'm just going to go back to my account and then copy over the token again. So, what you want to do anytime you find like a less than token, greater than symbol, or like your API token or something like this, you just want to replace the entire thing with the token. Okay? You don't have to do it in between the little less than or greater than symbols. I used to run into issues like that all the time. Then the last thing is, if you think about it, like we've worked through all of the information here. We have the actor. We have the endpoint. We have the token. We don't have any input though. Like we're not actually feeding it any data. You guys remember how before if we go to the LinkedIn job scraper, we fed it this job search URL. Well, now we're not doing so. Well, the way you do this in Appify is just go to the JSON tab and we actually get all of the data in JavaScript that you can then just go back to NAN and then feed into using JSON just like this. So now we're sort of feeding it in everything including all of the settings and whatnot. What I'm going to do is I'm going to click execute step and I'm going just cross my fingers and hope that this works. And if it doesn't, we'll deal with it then. Okay. And then a quick and easy way to verify that this is actually running is just go back to the runs tab. And you can actually see it go and accumulate results for us live. We also obviously have the other data like the usage, how much money we're spending and so on and so forth. because this is like 100 or something uh sorry 1 cent per 10 I believe like we can just do the math and 78 cents 78 or 078 and then this is 78 results. Okay and it looks like we just wrapped up here. So this is in table format uh format. I'll put this in schema just cuz that's a little bit easier to see. And we now get all that data that we got in the app dashboard but we have it inside of NAN which is obviously much more powerful because NAD is the glue that allows us to connect with a million other platforms. I'm just going to go back to our road map here. Let me make it a little more convenient for me. And I'm going to check this just cuz I always like to have some sort of record of progress. And then I'm going to check this as well. And now our next task is to compare incoming jobs against some bank of skills and have AI filter. So the question is how do we do that? Well, first of all, what I'm going to do is I'm just going to pin all of these outputs because I don't want to have to rerun that. You guys see how that took me like a solid minute or two. I was just waiting around. Maybe you guys didn't cuz we cut the video or whatever. But imagine if every single time you wanted to test a flow, it took you a minute. You know how many times I test a flow? I test a flow like 100 times per build. So that's an additional 100 minutes or hour and a half. Just cut all that out by pinning data so you never have to do any of those weights again. If you do incremental testing like I do, this saves you a ton of time. And incremental testing in and of itself saves you a ton of time. So we just highly recommend anytime you do anything like this. But anyway, the next question is we actually have to do some filtering, right? So why don't we do some filtering? How am I going to do it? I'm going to feed this into OpenAI. I'm going to go message model. And there's sort of like two ways that we could do filtering like logically. Think about it. You can do filtering procedurally. Procedurally just means looking at keywords and stuff. Like that's like kind of the old way doing filtering. Then you could do filtering flexibly. And this is where you know even if something maybe doesn't have the keyword you're looking for, you feed into AI. AI can like infer that you know even though it's not the same word, it's like a word that you're looking for. It's more of a concept based thing, right? So instead of using keywords, we're going to do this. But I want you to know that you can totally still filter based off keywords. Like you could make it so like you only apply to jobs if it has the term NAD in it because you're an NAD engineer or something, right? We're just going to use AI here just cuz I love using AI specifically for this purpose. It's just a lot more flexible. It's a lot easier to use. Okay, so scrolling all the way down here to the bottom. First, we have to create a credential. If you guys haven't done this, just check out the documentation. They actually show you how step by step. You basically just open your API keys page, create a new secret key, and then paste it inside of NAN. So, I've done this before in multiple videos. I'm just going to leave that out now, but this is where you put it. Then you click save and all of a sudden now you're connected to your open AAI account which is great. We're going to use the resource text operation message model. And from here I'm just going to use GPT 4. 1 and go mini. And what I'm going to do is I'm just going to write a big list of skills basically that I have. And then I'm going to ask it, hey, if the job doesn't include information that you feel is pertinent to me, then just don't let it proceed or something. Okay. So, first things first, we're going to go to system and I'll say you're a helpful intelligent job filtering assistant. I'm going to go user and I will say you filter jobs based on a list of attributes and skills that I or let me give it some uh context. I'm looking for jobs. Your task is to filter them based on a list of attributes and skills that I have. some jobs may not be relevant, which is why I want you to go through each of them and then let me know whether or not I'm an okay fit. Below is a giant list of all of my skills. And then underneath here, I'm going to say here is the job description. Then I will say respond in this JSON format. And I will say verdict true or false. So now it's going to return true or false depending on whether or not I'm a fit. I'm a fit, return true. If I'm not a fit, return false. both strings. Let's just say that make things a little bit easier for me. Then we will go output randomness temperature. And I always do 0. 7 for applications like this just because I want it a little bit more deterministic, a little bit less random. Okay. So now what I'm going to do is I'm going to give it just a bunch of skills. Maybe I'll just give it like context. Below is a block of context about me and my skills. Let's do that. And then here is the job description. Let's start with skills. I'm just going to go and have AI whip up a big thing for me. Okay, what I did is I just had AI do a little bit of research on me and then dump a bunch of context. Now, what I'm going to say is job description wise, all I do for this is I actually go expression and then this is kind of a hack you guys could use if you don't know what to feed in or if you're just lazy like me. AI is at the point now where um the context window which is like the amount of tokens that you feed in the number rather is so long that you could feed in like 20 bucks and you'd still be fine. Maybe not 20 bucks. It depends on the model. But because of this tokens are very cheap and so inference is cheap. And so what I do now is I just feed in the entire JSON object. I no longer deal with like trying to make this perfect. I just go JSON to JSON string and then I feed this entire thing in just like this. Okay. So once I'm done with that, we're actually good to go. So, I'm going to open this up now. And then, do you guys see how it says 100 items here? If I just ran this as is, it would take quite a while to finish. So, what I'm going to do is I'll use a limit node. I'm just going to feed in one item at a time. And that's just going to kind of be my hack. Okay. So, we're going to execute the workflow. We're just going to feed in a single item past here. And we're going to run it. And because we had a very simple and straightforward and easy pitch, the verdict is just immediately false. Like, it wasn't really that big of a deal. It wasn't very long, right? Okay. So, what I'm going to do now is I'm going to pin this. I'll also pin this. What do we do with this data? You know, it just said verdict false. We can actually just feed this into a filter now. And we could say, you know, verdict has to be equal to true. And if it's not equal to true, then we don't proceed. So, let's execute this. And now, you know, we discarded verdict cuz it was false. So, we're just not going to use this anymore, right? And so, this is the design pattern that you use anytime you have AI doing some filtering. You drop an AI model that has like a returns JSON true or false. You just have a filter after that just checks to see whether it's true or false. So, that's pretty good to me. Why don't we now feed in 10 items? Let's unpin all of these and then we can go and we can see what AI has to say about all 10. Obviously going to take significantly longer because in N8 we run all 10 of these in parallel. So it's actually just like all 10 of these are going on. And it looks like of the 10 that we fed in, it spit out five. So let me just take a peek at what five it spat out. Well, I guess I don't actually have too much access to previous data, huh? That's unfortunate. Still, that's cool. Let us just very quickly pin this over here. Just feed in Jason. Let's see. The first item we fed in was this one here, which is vice president of data. Looks like the verdict was false. Looks like the second one was looking for AI product manager and business analyst. That one was true. That seems pretty reasonable based off what I'm looking for. It looks like the third one was senior data scientist. That was false. Probably because I'm not a data scientist. This one here is true. Anyway, I'm just going to trust that the filter is doing what it's doing. I don't do this because I actually have a big list of criteria or whatever. Obviously, the purpose of this live build is to show you guys how to build something like this that you guys can then use at scale for whatever job application process you guys want. So, just treat this as like a nugget and then go in and like add your own skills, add your own contacts, add everything about you, and it'll do a good job. From here, what do we actually do? Well, let me think. We go back here. we can check it. Now what we do is we just have another AI model that customizes our resume to match the specifications and standards that we give to it. So basically we have to do now is you see this we have to feed this in to another AI call and that plus the data in the HTTP request. It's just going to kind of like meld together in this lovely open AI call. Then from here we're going to spit out a customized resume and then we'll be done probably about half of our build already. So, as you guys could see, pretty straightforward, pretty standard. So, what I'm going to do first then is I'm just going to copy the OpenAI model, paste it in here. Then, the reason why I do that is, you know, I just have most of it set up for me already. Then, instead of GPT4 Mini, which I think does an okay job, I'm just going to use GPT4. 1. I think that, you know, the writing is a little bit more important than the simple filtering, yes or no. I'm going to say job customization assistant. Leave the I'm looking for jobs. I'll change to your task. Your task is to customize a provided resume. and then using a provided job description. So I'm going to leave this. So here's the job description and here I'm going to say here is my resume over here. We're now going to feed in the entire resume. Then I will say don't respond in JSON format. Respond with only the resume, nothing else. And I'll also say this resume will be added to a Google doc. So, write it in HTML format that I can easily copy and paste. Okay, I think this is going to be fine. I'm not entirely sure. We're going to give it a try, but this should be okay. Do we have everything we need? Job description. Yeah, I mean, I guess we do. It's not going to be JSON. No, it's going to be from this filter. It's actually HTTP request if you think about it. So, what is the job description? is going to be dollar sign and then where am I going? I'm going down to the HTTP request. And then I'm uh maybe I'm going to limit. Gez, I don't know. I feel like I could select from either of these. Let's go limit. And then I'm just going to convert the entire thing to a JSON string. Item. json and then to JSON string. There you go. Should now have everything. We'll see. I think this is going to work. I mean, it's going to be a little bit different. Looks like for whatever reason there is a space with two hiring manager. Let me just go back here and make sure I include a space. I guess I did, huh? All right, whatever. We'll leave it as is. It should be able to do a pretty good job here. So, I'm just going to unpin this. Uh, this is five items. I just want to test it again on one. So, I'm going to go limit for now. I'm just going to do one, and I'm going to execute this workflow. Now, that's going to immediately execute all this. Uh, looks like using the item method doesn't work with pin data. Please unpin filter and try again. So, I think we're going to have to unpin a couple of these, which is unfortunate. Let's try this. Yeah, same deal. I think we got to do the basically everything from here, which sucks. Got to filter this out. Feeding it 10 items, right? Okay, there we go. So, sometimes when you use the item method, so when you go limit item dollar sign JSON or whatever, you run into what are called item matching issues in NAD where it doesn't really know which one to reference because you've pinned some data but you haven't pinned others. So, a quick and easy hack is just to like unpin everything except for whatever super crazy API call you need and then you're good to go. So, uh we're outputting this with like back ticks. I really don't want it to output with back ticks. Yeah, I don't like that. I don't like that at all. Can I just say do not output any back text? No. this and I will say your first character should be Let's just try this. Execute this again. Is this going to work? I think so. Yeah, it's already running so it's probably good. Okay, let's take a peek at this. This looks to be keeping in a bunch of additional information that I don't want. Like this output a style. I think I might just ask it to go markdown format instead of HTML. Markdown ETX format. Let's do this. I'll just try this one more time. And then, you know, I'm just going to output markdown and then convert the markdown to HTML. That's a little bit easier. I think as you guys see I normally like using output content as JSON just a lot easier because then I can map the data directly and also anecdotally I found the quality of the output is usually a little bit more systematized. This looks good to me though. Okay cool. So now that I'm done with this what do we want to do? Actually need to u think about it like upload it to a Google doc. So first of all I'm just going to convert the HTML to markdown or markdown to HTML. Feed in this key. Execute that step. So now I'm feeding in the output. And now what we need to do is we need to create and then update a Google doc. I don't know if we can update it with HTML though. Can we? Yeah. I don't know if we can. So why don't we do this first? Let's just create the Google doc. This is like our next Google doc. We're going to call it Nick Sar. Is this the best thing you could call it? No. You could actually call it some else. You know, you could say like foris, you know, whatever the name of the company is. But for now, we're going to create the resume. What we have to do after we do the creating of the resume is we actually have to like update the resume. Now, I think the only way to actually do this, and I only know this cuz somebody in Maker School mentioned this earlier, is, you know, Google Docs has this way where if you just apply HTML, it'll actually just automatically create the doc with nice formatting and stuff like that. But, NAN doesn't have a built-in endpoint for it. So, basically, you have to have do a custom HTTP request. So, I'm going to kind of cheat here because I know how to do this and I'm not going to like work out the logic with you as much, but I've actually done this before in a previous build. So, I'm just going to scroll through my builds and I'm going to see, can I find the thing where I make the Google doc because, you know, I've done it before and it's pretty cool. Okay, so I did some digging and I found this HTTP request. Here's how it works. Use what's called a patch update. You feed in this URL here. So, googleapis. com/upload/drivev3/files then the ID of the document that we just created a moment ago. Upload type equals media. Then what you do is you feed in text/html with the body from this markdown node. So I'm going to go and grab all of this data. And this should now update the doc that we just created a minute ago. in order to do authentication with like let's say you have a situation where you have a built-in like the Google Docs builtin that we just used to create the doc but then you have to do some sort of HTTP request that is not supported by nadens builtins there's a quick and easy hack that lets you do it way easier than what most people are probably doing which is just rebuilding everything and knocking around with authentication if you go to authentication type predefined credential type here then under credential type actually just put in the thing that you connected with in the previous built-in module so in my case Google docs oth2 API This will actually automatically pull the same connection that we just created earlier except they'll put it in an HTTP request. And they do this to simplify you having to stitch together platform integrations and then custom API requests like we're doing now. So very useful. Okay. So after I've done this, what I'm going to do is I'm going to pin this because I want the same ID to come out. See here it says Nixive Resume, right? I should actually be able to go back here, go docs. google. com google. com and see nyx resume again. Cool. So, it's actually right over here. You guys see how it's totally empty right now? Well, it's totally empty cuz all we did was we created the document. We haven't updated yet. So, what I'm going to do is I'm going to see if this works to update. So, I'm going to say add HTML to let's resume here. Execute workflow. Nice. Awesome. And now I actually have my customized resume which takes all of the information that I used and gave it initially. Then it weaves in what the job is looking for alongside my experience to rewrite me something that looks like this. So that's pretty neat. Does a pretty good job here. I'm a business focused AI product manager and automation specialist with six years of hands-on experience designing, deploying, and scaling intelligent automation solutions for SMBs and SAS. Proven track record bringing structured ambiguity, translating user pain points into higher ROI product requirements, and bridging business and technical teams. I mean, that's pretty great. And to be honest, that's more or less exactly what I do. So uncanny. And then it also did formatting here with like my email address and stuff like that, which I like. And yeah, the formatting looks pretty clean. So if you think about it, like we could provide it some more, you know, formatting things if we wanted it. We could say, hey, make sure to output the chapter headings as like H2s or H3s or whatever. And it would do a pretty good job. We could also give it some examples of what we think really high quality rums are. So we can do that transformation by looking at um, you know, our examples, not necessarily just our oneline instructions. But all in still a pretty good job. Um, and I like where this is going. So, I'm just going to pin this now and then we are going to just check this puppy off. Last thing is we need to enrich the website and any other information for contact details and then send an email using a short punchy pitch and attach a resume. Okay, so what do we have now? Just taken stock. We have a very long linear flow that does Google doc content extraction. We call a specific job search URL. We then filter using AI. Let me just update this now. So, it says filter job. Then we do some math over here. or some funky processing. Then over here we rewrite the resume. Over here we convert it to markdown. Over here we create new resume. Then here we add HTML to the new resume. What's going on after this? Well, if this is all like data wrangling. Well, everything on the right of this is basically now just outreach. So it's sort of like a classic outreach flow. Now, if you guys have been following my channel for a while, you know that I do outreach flows all the time. What I'm going to do for this outreach flow is I'm just going to take the website. Then I'm going to see if we can enrich it using a service called any mailfinder. Enrich in this case just means grab decision makers at these companies. And then once we have their emails, we'll just feed that into a little email template and then send it out, including our attached resume. Nothing super fancy here. This is just an outreach flow. And uh you know, if you guys have seen me do this thing before, you'll know that outreach flows, even simple ones like this, can be extraordinarily powerful in the right hands. So, I'm going to use this service here called any mailfinder. There are a million services you could use. Okay. But I like any mailfinder because any mailinder allows me to find any mail. No, it's just a good service. It's reasonably cost effective at like higher levels and yeah, these guys are just like pretty solid. So, how do we use it? We use company search, person search, decision maker search, LinkedIn URL search. If I type leftclick. ai in here, I look for the CEO or owner, it actually goes and it finds me. plus my email address. Pretty wonky, huh? Isn't that crazy how it just does that? So, imagine if I feed in some other company, you know, zapier. com or whatever. If I feed it in here, it will find me the email address of the decision maker here. And so, if we constrain the company size to something small like 50 to 100, maybe like 1 to 100 or so, odds are the founder or decision maker that we find is either going to be equivalent to the person doing the hiring, like it's going to be them, or they're just going to be known to them and can very easily forward over our inquiry or whatever. And odds are if somebody's a business owner or a founder, decision maker at a company that like, you know, has high agency and makes things happen, when we reach out to them with something that's customized and punchy like this and we put proof in the pudding and we show them like, hey, we actually went the extra mile. We're not just like applying to a job with a crappy resume. A lot of the time they really appreciate that cuz they're the sort of person that respects that thing cuz they're entrepreneurial in general, right? So yeah, that's sort of like what all this banks on. And I found great success using methods just like this for businessto business outreach, entrepreneurial, and I do it all the time. So that's what we're going to do here. I have 65,000 credits, so I might as well start working through them. Got an API key over here. So, I'm actually just going to copy this and then I'll also read the API documentation while I'm at it. What do we want? We want decision makers email. Looks like we do this with any mailinder v5. Guess what? We even have a curl request over here. So, I'm just going to copy over this curl request. And then I'll go back here. And now I'm just going to do another HTTP request. Import curl. Paste all this in. Done. Just need to fill in my API key, which we just got a moment ago. back over here. Right. Copy that. Paste it. What else do we want? Domain Microsoft. Decision maker category CEO. Well, if I just execute this, can I get the CEO of Microsoft? I don't know. Let's give it a try. Yeah. Now, I think his email is basically public knowledge at this point, but you guys get the idea, right? If we do leftclick. AI. Are we going to find anything? Let's see. We may. We may not. Nice. Yeah, we found me, right? That's kind of neat. So now that we have this, I'm just going to feed it in one at a time. And what are we going to feed in? Well, we just need the website URL, right? So I think what we want is we want this limit node. And then what we want here is we want the domain, want the website, right? Maybe I'll just type in a website. I'll find it. Man, this is really long. Okay, right over here. Company website. Cool. Now, we're not going to be able to get this because we're going to run into that like node pinning thing again, but that's okay. There's one more thing I want to do here. There is the chance, and I don't know for sure, but I'm thinking there could be a chance that we feed in a record that doesn't have a website. So, I just want to go back to my filter node, and I basically want to not only filter for the verdict, but I want to make sure they have a website. So, I'm going to go back here to that limit node, and then I'll go website again. Scroll down just to where it says company website. I'll basically just say this is not empty. Okay, so we actually have a value for the website. Only if that's true are we going to proceed. So true and website. Let's just like filter true and website limit two. Okay, now that we have all this, I'm going to grab everything. Press P. Call this get email. And then we'll only proceed with this assuming that actually find the email. So say only proceed if email is not empty basically. Okay, which I think is good. And that way you know like email present at the very end we could do some sort of like Gmail flow. Anyway, we'll do that in a sec. So sorry, let's keep these pinned and then let's just run through the whole flow one more time. With 10 jobs, let's see how many of these emails we can get. Realistically, we usually get between 30 to 45%. So it's not a sure thing, right? Like not all of these have actual email addresses associated with them obviously, but still we can get reasonably high enrichment rate I want to say. So that's what we're doing. Kind of looks like we had an issue. The resource you are requesting could not be found. So no underscore result. I don't want it to return me an actual error. That's kind of dumb. So instead of returning an actual error now, hold on. Is this because we're feeding in the www? H I don't know. I don't actually know. There could be a couple of errors here, right? Like I'm going to pin this now. Just got to run this. Oops. Sorry, not uh don't pin this. If you think about it, if I execute this step, it runs. This is a resource. It cannot be found. What if I just copy this and then I paste this without this? You know, sometimes there's a difference between like domain name and then resource. No. Okay. So, I think this is just like a basically if this doesn't find the email just returns an error. So, what do I do on error? I actually want to continue using the error output. So when you do that, there's a success route and then there's an error output. So success route will maintain, error output will stop. So let's execute this. Now this error output should go down here. Cool. So now if we feed in more than one element, right? Like and three of them are good and then one of them isn't, it should be fine. So I'm going to go over here. Let's output this to let's just do three for now. H maybe we'll do four. Unpin all of these. And then I'm going to run it from here. And we're going to try and get at least one of these. Okay. Uh oh, good lord. Let's try this one more time. Yeah, the item matching issues in NAN are the big things, honestly. You basically just have to rerun this every single time starting from here, which is unfortunate. And I'm just getting some NAN connection issues because the server was down earlier. We do be doing some filtering. And then I'm just going to get kind of ahead of myself here and just start working on some of the logic of a Gmail node. So I'm going to go create a So we could create an email. I'm just going to create a draft for simplicity. We connect this really easily. You just sign in with Google and voila, you're done. Resource is going to be draft. And I'm just going to say like re and then I'm going to map the job title. Then I'll just say first of all, let's see if we can actually get the name. That'd be really cool. But I'll probably see if I could use the name in the email and say, "Hey Pete, how's it going? " Then assuming I can find it, then I'm good. Now, one thing you'll see is this is taking a long time. I think it's been like almost 45 seconds or so now. when you pass in a ton of items to a flow like this, nine runs them all simultaneously. And so you don't actually see any completion till the very end. So what I usually recommend is if you're going to be running this on like massive data sets, what I use is I use a loop over items node instead of just like passing it all in through here. Obviously, it's a trade-off between the simplicity of an operation and then, you know, the scalability of an operation. I'm the sort of guy that just likes making things as simple as humanly possible. Running it, making sure that works, and then worrying about touching it up and making it nice and sexy after. But I think that's worth considering if you guys wanted to scale this up and do 100,000 of these simultaneously. Then that's what that would look like. Anyway, now that we're done with this, you can see we've actually have two successful items. I'm going to feed this in to this over here. And I'm first going to start by going all the way up to is it limit two? No limit, you know, re what's the job post, right? VP data and analytics. So, I'm going to say revp data and analytics. Do I think that this is the best prompt you could use? No. But I think it's still pretty reasonable. Then, if I go back to get email, maybe email exists. You can see there's a person's name here. So, I'm actually just going to feed in this person's full name. I'll say hey ud but in this case what I need to do is I need to split this into two based off the presence of a space and then I only want to get the first element and so this is a quick and easy way to grab I think there's probably also like a first isn't there yeah there's a first I'll use first that's simpler for you guys so the message is now hey then underneath here I could actually like tell them that I just wanted to let you know I'm the right fit for the job. Resume attached. I just want I actually used AI to scrape. I prefer showing to telling. So to be upfront, I actually use AI to scrape this job, customize my resume, then automatically get your contact details. We'd be happy to run you through the system. And I think, you know, we can actually feed in the name of the company. Well, I don't know. Should we? know you're hiring right now. Just wanted to tell you I'm the right fit for the job. To be upfront, a to scrape this job, customize resume, and then automatically get your contact details. I know showing is better than telling, so would be happy to run you any hiring managers here through the system. I plan on going above and beyond. Yeah, if picked, I'd go far above and beyond the job description. Blint systems similar to this in We probably should grab the name of the business. I kind of dumb not to. Let's say company name right over here. Awesome. Okay, then we obviously need to make an attachment. What do we have to do? Well, we actually have the Google Doc, right? So, actually, if we're doing Gmail, we can just feed in the resume um by linking it. So, that's probably what I'm going to do. You get the document ID here. How you link a Google Docs, you just grab this, feed the URL in, and right over here, you just feed in the document ID. See that? Then we go back slashedit. And I'm just actually going to feed that in as is. Let me just see if I just remove this. Paste this. Does that work? Okay, cool. Now, what if I make this visible to anyone with the link? Does that work? Yep, that works. Cool. So, now I'm going to feed this in. And then we should be good to go. I don't know for sure how this is going to work. if this is going to work. I'd like this to work. So, all I need to do is just do this and I create a draft. I don't know if it's going to know to select the right record. Yeah, it's not going record, unfortunately. So, we do have to rerun this. Um, for simplicity's sake, why don't I do eight over here? And then over here, why don't I just do three? Save this. Oh, and there's one more thing I need to do. Um, creating a doc is not enough. In addition to creating a doc, what we have to do is we have to make it sharable. So there's a simple and easy way to do this here with Google Drive. We just share a file. Click this button. The file that you want to share, you just do using an ID. So this is what you would feed in. Then over here, the permissions you just make everybody a writer. And then that would allow us to share the file without actually having to do anything different. So, I'm just going to space this out now. I'm going to execute this one more time. Just while this is executing, I'm going to make this a little bit sexier, or at least more fitable into, you know, like one line here, just going to make for a better screenshot. Let's move this down here. That should be good. The reason why we're getting this is we probably referencing the wrong JSON ID now. Yeah, we are. We got to go back to create new resume. Let's fix that up. by dragging in this to file slash feed that in right there. Cool. And then also while this is running I'm just going to go and create some notes. So let's go sticky note top right hand corner we will say AI automated resume system and then what do we actually have to update? Guess we just have to use the any mail finder key. Update your resume template ID in the Google Docs module and your any mail finder API key in the uh your appy app module and your any mailinder API key in the any mailinder module. There you go. Okay, cool. Now, we've actually queued up an email. So, I'm just going to head over to my email address here and go drafts. And as we see, it says reai transformation principle. Hey, Udy, I know you're hiring right now. Just wanted to tell you that I'm the right fit for the job. To be upfront, I actually use to scrape this job, customize my resume, and automatically get your contact details. I know showing is better than telling, so be happy to run you and any hiring managers here through the system. If picked, I'd be going far above and beyond the job description, and implement systems similar to this in Cyber Arc. Resume below. Let me know what you'd like to do next. Thanks, Nick. And that's it. In a nutshell, pretty straightforward and I think it's clear at the end of the day that stuff like this works, right? You're putting yourself in front of decision makers and people that ultimately are interested in seeing high agency people. And then when you combine that with the fact that we are applying inside of an automation, which is not something you have to do, but it's something you can do. Obviously, you guys can see the benefits of that. And
