#209: Transforming the hiring process with JobHive

In this episode, we talk with Aaron Jorgensen about how JobHive came to life - starting as a small résumé-parsing experiment and gradually growing into a structured, AI-supported interview workflow. Aaron explains how the system handles voice capture, transcription, prompts, and AI avatars, and why he moved toward a multi-agent approach instead of relying on one model to do everything.

We dig into what “fair scoring” actually means, how cross-checking evaluators and confidence levels work, and why it’s important to keep the reasoning behind decisions visible to both employers and candidates.

From the builder’s perspective, Aaron walks through the practical side of developing the platform: shaping an MVP, working with LangChain, choosing AWS tools that reduce overhead, and dealing with the usual setbacks—broken features, unreliable external services, and the moments that test your patience. He also talks about the routines and habits that helped him stay consistent during the harder stretches.

If you’re interested in hiring workflows, AI tooling, or the reality of turning a rough prototype into a functioning product, this conversation covers it all.

To learn more about Aaron's work, check out his websites or reach out to him on socials:

JobHive: https://jobhive.ai

Aaron's Website: https://ajeema.com/

LinkedIn: https://www.linkedin.com/in/mraaronjorgensen/

Circle: https://pybites.circle.so/u/22287446

___

Book mentioned in ep: https://pybitesbooks.com/books/P3EFa-WuMMkC

___

💡🧑‍💻 Want to become a more focused, motivated, and effective Python developer? The Pybites Developer Mindset (PDM) Program helps you build strong habits, deepen your skills, and make real progress in just six weeks. Join us for mentorship, accountability, and a supportive community that keeps you moving forward. Start your PDM journey now! 🌟✅ https://pybit.es/catalogue/the-pdm-program/

___

If you found this podcast helpful, please consider following us!
Start Here with Pybites: https://pybit.es

Transcript

Minimal Viable Mindset

Aaron 0:00

That's very important when you're building anything. It's uh you know how can I what's the minimal level product? What is the simplest thing uh that I can build to at least remote one step while keeping your bigger picture? I knew what I wanted to build. I didn't necessarily know how great it would be from and all the things that would add on as a build, but I knew where I wanted to get to, and then I kind of stepped back and tried the most basic stuff first.

Julian 0:28

Hello and welcome to the Pivites Podcast, where we talk about Python career and mindset. Later hosts, I'm Julian Takara, and I am Bob Belderboy.

Bob 0:37

If you're looking to improve your Python, your career, and learn the mindset, this is the podcast. Let's get started. Hello, welcome back everybody to the Pivites Podcast. It's Bob and I'm here with uh Aaron Jorgensen. Hi, Aaron, how are you doing?

Aaron 0:53

Hey, I'm doing great. How are you, Bob?

Bob 0:55

Yeah, great. Um happy we finally uh get to meet on the podcast again. Because we uh we've of course been in touch for five years because you're one we're one of our first PDM clients, and uh here we are like five years later. So I'm happy that you're back.

Aaron 1:10

Long time coming, excited.

Bob 1:12

Yeah, yeah, because we had this in uh planning for like three months and finally going to do it. So uh yeah, uh today we're going to talk about your story, uh Job Hive, an exciting platform you're building. Uh the mindset of a developer, of course, as always, and uh yeah, maybe some workout stuff, uh definitely mindset books. But yeah, centering around job hive. So, but before we dive into it, do you want to give a quick intro to the audience? Who you are and what you do?

Aaron 1:41

Sure, yeah. So my name is Aaron Jorgensen. I've been in tech for what feels like my entire life uh for a very long time, in and out of various roles. Um, but you know, they're not always the same. I've been from a site reliability manager, I've played some CTO roles, you know, I've done a lot of different uh uh roles and more into lots of different hats, but it's all driven by a passion for technology. Currently building something on my own that I'm very excited about, we'll talk about today. Uh some other fun facts. I have a little hobby farm. So we uh we have a bunch of goats and chickens and ducks, and gosh, I lost count of everything, but we have a lot of little animals running around constantly. Uh we are live, I'm living currently about five hours north of Vancouver, uh in a more rural kind of area. Um I have a lake view in front of me. It's not uncommon to see a deer on Octoping literally walk up and put its face against my window. So it's uh it's a beautiful life.

Bob 2:42

That's the dream. I love how you uh combine coding with uh rural live and you're building houses and and it's it's amazing. Yeah, cool, cool. Do you have a win of the week?

JobHive Overview And Core Value

Aaron 2:54

Oh, I got lots of wins of the week. I would say I like a lot. Um one would be probably I'm in the middle of training an AI model for Job Hive, and I'm running the second pass, the first pass I finished yesterday, and so far, so good. It's uh it's quite impressive. So I'm quite excited about that. Yeah, nice.

Bob 3:19

We're definitely gonna dive into that. Yeah, but before we go into the nitty-gritty, um, yeah, tell us about Jobhive, the exciting platform you're building. Um, what's the the problem itself? Where did it come from? And then we can go deeper into the stack and all that.

Aaron 3:34

Yeah, definitely. So JobHive is an AI-powered uh interview platform. It uh allows you to do mock interviews or to do real interviews, and it provides you and the employer detailed insights on how the how they did in the interview, everywhere from like a sentiment analysis to pace of speech to red flags or inconsistencies, uh, job matching, culture, um, behavioral analytics. So very detailed and uh provides both, you know, like I said, the employer all the details, but also the candidate all the details so they can know where they stand, how they did, and also areas for improvement. There's a there's a whole section that we do where we talk about learning gaps and recommendations that we provide that to the candidate. So not only do they see how they did, but they also get ideas and how they can strengthen, you know, to do better in the future.

Bob 4:35

Nice. So um the customers, both the interviewer as well as the interviewee?

Aaron 4:43

Yeah, so we cater to both candidates and employers, employers being the the bulk of you know who our our uh customers would be, but you know, we we kind of cater to both, and uh yeah, that's what we do.

Bob 4:58

Cool. And how how does it work? Um, the the process like um well, we get into the model training, I guess, in a bit, but from a UI, UX uh perspective. Um, how how does the recruiter, for example, set it up and where's the AI involved? How does it work in more detail?

Origin Story: Resume Parsing To Interviews

Aaron 5:17

Sure. So the I didn't quite start about the origin story. So I'm gonna I would like to go back slightly, explain the origin story.

Bob 5:25

Yeah, yeah, the do the origin story first. Yeah. And he kind of paints an exciting picture.

Prototyping The First Voice-To-LLM Flow

Aaron 5:29

So back the reason it all started was uh in one of my previous roles, every year we do co-op hire, and every department gets uh you know a stack of resumes from the local universities of students that need some co-op. And you know, they usually get to kind of see what area they like. And so I got a stack of probably a hundred resumes of co-op students that are interested in reliability and observability and and stuff like that. And so I'm I'm looking at the stack and I'm I'm like, I don't have the time to go through a hundred resumes on top of everything else I'm doing. And and that was the case for every department, all of my peers, they all have the same thing. And it's exciting to have a co-op student and to be able to do that process, but it it takes a lot of time. So I thought, well, why don't I write a script that can quickly parse through all of the resumes and and find the top 10, 15 that match the most of what I'm looking for? And so I did that. I wrote the script, it parsed it, it was fantastic. The top 15 that I I got out of it, I went through, and they were definitely the top 15. I spent a little bit of time digging through just to verify if I missed any. Uh, and it did a great job. So I actually said, well, anybody else want to use this and kind of pass it around as, you know, let's share the let's share the benefits and reduce the the toil. Uh anyway, so I did that, and then I was set up the interview. So now I have 15 interviews to do. And you know, I don't I love talking to people, I I don't mind doing the engagement, but after the 10th one, I'm you know, you're asking the same questions. It's redundant to you, it's new for them, right? So you have to respect their time. They've never done this particular interview before. I'm doing it for the 10th, 15th time. Uh, and so you know, I got through that whole process and I just couldn't shake the idea of there's gotta be a better way, uh, you know, that we can we can do this while still making it a human-like experience, something that's still unique to them, something that still provides value, but isn't this time you know intensive? Uh and so that kind of just sat in my head, and we didn't went to the whole co-op as normal and you know, and time progressed, but I couldn't get it out of my head with this this uh this idea. And so I sat down one evening and I started you know playing around with it, and I was super happy, uh, you know, and I thought, well, let's just see where I can get it done, get done. And so I sat down and I was for evening, and I told my wife, I'm just gonna sit down, I'm gonna play with something that I can't get out of my head. Uh and so I sat down, I spun up a Django uh stack uh Bing Python back end, love it. It's you know, it's fantastic. So I spun up a Django backend and a super, super hacky uh HTML front end where I would just the whole idea was all it would do is I would go on and use the mic browser to you know to record my voice, it would save the audio file, and then now what? So that's great. That's you know the first step. And then I thought, well, let's take this audio file and put it in the text so I have something workable for the LM. And this is before kind of a lot of these multi-models were out where you know can you receive audio input? And so I I transcribed it with with Whisper, sent it over to OpenAI, said with a really cheesy prompt, this is you know, this is an interviewer, this is a response to something. What would you say in return? Kind of just to get a dialogue back and forth. And you know, I got the LLM response. I'm like, well, now what do I do with this? So essentially that's how it started. It was very hacky. An idea that I couldn't get out of my head, and then as we'll get into it, it it's become what it is now. Amazing.

Bob 9:26

Yeah, when when was that more or less?

Aaron 9:29

How long ago? Uh about a year and a half ago now.

Bob 9:34

Yeah. So ChatGPT was around, it was already probably a year and a half, maybe two years.

Aaron 9:42

So chat GPT was around, you know, it was kind of the the newer shiny thing. Uh a lot of vagueness around how to prompt, you know, we weren't as great as we are nowadays. I mean, it's crazy how fast things are changing, right?

Bob 9:57

So yeah, yeah.

Aaron 9:58

But back then it was, you know, just whatever like straight text, just tell me how to respond to this kind of thing. No, no great prompting or anything. Yeah, but I guess we got going.

Bob 10:10

Yeah, no, amazing. I mean, what stands out is that you build it first and foremost for yourself, right? You had that need, you saw that opportunity for automation, and when there's a sign of automation, then you need to build a tool, right? If you're a developer, that's the natural thing to do. And you have the same thing, you just get obsessed about it, and you just have to get it out at some point, you know. There's like every day, not right.

Aaron 10:34

It's always uh the model I always go by, and it's automate the boring stuff, right? So yeah, I'm always what can I automate? What can I how can I reduce toil? And I apply that in in all of my previous roles as a cell reliability manager, but it's it's just that mindset, right? Like automate what you can so you can focus on the more important things.

Bob 10:52

So yeah, then you sit down and you start to solve the simplest problem. Well, not the simplest necessary, but just the most straightforward or the most obvious first step, right? Which was to record your voice and then to transcribe it and then send it to the model, which of course is maybe 10-20 percent of where you have to be. But that critical step, doing it, accepting that it's hacky, that was is what you get you to the next step, and then it starts snowballing, right?

Aaron 11:18

Right, right. Yeah, I mean, and it wasn't even wasn't even automated. I'd I'd have to you know record the voice, it would go to an audio file, then I'd I'd manually try, you know, run a scripts in Python to you know transcribe it, and then I'd use another. And so it wasn't like there wasn't a beautiful automated pipeline that triggered it off every flow, but I knew what essentially I needed to accomplish, right? Based on doing interviews, you need to listen to the person, you need to respond to the person. Uh, and so I started trying to figure out based on that, you know, what are the various small basic steps just to verify something. And that's all that's very important when you're building anything. Is uh, you know, how can I what's the minimal value of a product? What is the simplest thing uh that I can build to at least prove out a concept while keeping your the bigger picture? I knew what I wanted to build, I didn't necessarily know how great it would become and all the things that would add on as you go, but I knew where I wanted to get to, and then I kind of stepped back and tried the most basic steps first.

Bob 12:21

Right. And and at that point, what did you still need to add for it to call it MVP? Because you got a response, you might you might have the computer say it out loud, that's not too difficult, but I guess you would have to have some sort of conversation with multiple interactions, right? Was that what you then build next?

Moving To AWS And LangChain Orchestration

Aaron 12:39

Yeah, so what what I what my idea of success, so to speak, to have my minimal viable was that I need to come to something, a web page or something, and I need to have an engagement. So I need it to hear me and I need it to respond. So, you know, it's still at that point, it was the next step was it was still uh you know, pushing a button to record my voice, but then you know it would do the automated part. I figured out the you know, the web socket and the whole connections and everything, and then it brought it back, and then it would bring it back as text. So even that wasn't you know the the most fancy of things. Obviously, you'd want a voice coming back, but I figured let's start with at least a text being displayed back to them. So it was like a full flow, and then and then from there I started um thinking, well, now I need to get a voice into it. So I I leveraged Langchain, which is a very, very powerful and great you know orchestration tool for LLMs and agents and so on. So I grabbed, I grabbed down and I wrote a page, uh a script that had Langchain that would con and Langchain has a bunch of built-in connectors to various things, like you know, AWS bedrock and and various things. So that simplified. So where where you can use powerful libraries, use powerful libraries, don't rewrite everything. Uh and this was a very great library. So uh I then I connected it to AWS for the for the uh I started abandoning OpenAI because I figured I'd like to keep everything in in one area, and there I was already planning on hosting everything in AWS, so I said, well, let's let's look to you know how I can connect to AWS native tools. Uh and so the AWS has transcribe. So I hooked into transcribe to you know to change the audio to text, and it was that flow was you know, the audio file goes up, it goes over to S3 bucket, which triggers a lambda, which triggers the transcript. Like it wasn't the fastest of flows, but it worked. And uh, and then the next step was to tie it into AWS Bedrock, their LLM inference models. And then then they have uh also a speech model or a uh text-to-speech called Tali. So it I started you know seeing what tools AWS had, and and because Langchain is a great orchestrator, I started trying to plug it in and eventually got a web page where I could, you know, I still had to push a button, but I push button record, and then everything else would happen. It would, you know, send it up the audio file, transcribe it, turn it into voice, and then play the voice back to me. So at that point, I you know, it's like, okay, well, I I have the minimal viable products, I've proven out it works. That the LLM responses weren't fantastic. Sometimes they would hallucinate or they would talk about, you know, easily be sidetracked and so on. So it wasn't the most natural conversation, but it was a conversation. And that kind of set a fire and excitement in that okay, I got I got this going. Now it's fine-tuning, now it's tweaking, now it's building um the framework and logic works and and you know, and it's exciting. So then it just kept building from there.

From MVP To Product: Feedback Loops

Bob 16:08

Yeah, that's fascinating. So then at that point, you're from script to aha, this could be a product, right? Because now you're you solved your problem. At that point, did you get more users in? Did you get more eyes on it? What what was the next step? I mean, apart from fine-tuning, what what happened then for it to be a very minimal viable product to the fantastic landing page you see today today, and and really it now you're you're launching it now with you know, you're in the launching phase, right? So uh but it looks looks amazing if you go to the website. So what what was the transition from there on?

Aaron 16:48

Yeah, so I mean one thing I would caution is you know, I would say it's always get feedback early, click a small feedback loop early on. Uh so I you know, I was asking people I knew, people in HR, you know, what are some of your pain points? What are some of your problems? What would be, you know, I just asked a lot of questions. I would go online, I'd go on Reddit threads, you name it, because I wanted to the most important thing to understand when you're starting a product or when you're trying to get somewhere is what is well, we already answered what my pain point was, and it's obviously other people's pain points, but is what what do they what other what do they need? What's the what do the customers want? And I think a lot of what happens is we as developers get lots of ideas, we take ownership of of what we're building, and we just keep going and we keep then we get to a phase where like, oh this might be a cool thing, and we keep building it, and we can get a very big platform that nobody wants. Right. And so I I found and before I even went into this a few years back, I was reading quite a lot of books about you know the Lee methodology and the whole thing about you know how Toyota had, you know, implement the Lee methodology and how you can um the whole pipe the whole process pipeline can be shut down if one errors um because you know of excellence you want to move forward. But but the uh the idea that you know you build very small and you build based on what the demand is, not what you're demanding them to take, essentially, right? So I asked a lot of questions. I asked a lot of people to look at it to give me ideas. Um and from that I started shaping, and then the ideas kept blossoming from there. And I think that's you know, the the benefit of where we got today is that it's a very powerful, very based system that answers a lot of problems instead of creating new ones by building something no one wants and then putting it in their face, and then you know, they just get confused.

Bias, Fairness, And Human In The Loop

Bob 18:54

So yeah, did you find that that feedback um was aligned with your initial problem, or were there some critical pivots?

Aaron 19:03

Well, I would say in in general, yes, they they they match the same idea of this takes a lot of time, you know, on off our schedule, it's a lot of repetitiveness. The the the pivot point for me was when I realized that a lot of the human interaction, and it's needed in in job hive, it's still in the loop. We still use human in the loop, and we'll talk about that. But a lot of the human interaction is bias in many areas, and not that we mean to be, but as I was talking to people, I noticed that there are human conditions when we're interviewing that affect how we hire, you know, and it may not be intentional, but it's there. And so I started to look at it from a lens of not not only do I want to create a platform and a tool that makes things easier and faster, but I want to do it in a way that is fair. And I want to do it in a way that there is no bias in in any of it. And we can we'll talk about how we get there. So I wanted to eliminate bias. And from that, there are these few key themes that I kind of pulled out in my interaction that made me pivot in the way that I built it out. Uh and considered at a very early stage, which shaped where it ended up, which was I think super critical. And what sets it apart now from anything else that's out there was I think considering outside of the box and these other things at the very onset, because my original roadmap was different in my head, what I had projected. And then having these conversations and thinking about it, and then really considering from the candidate's point of view, from the employer's point of view, taking the perspective of the eventual customers, pivoted uh and changed by roadmap to the better. I think.

Bob 21:09

Yeah, I think that's a critical point. Uh I I agree. As developers, we get all these ideas, and we just want to code and we think it's it's cool, but then nobody wants it, right? So sooner and the deeper you can talk with your customers or clients, uh, the better uh it will be for development and and steering the ship in the right direction.

Multi-Agent Design And Scoring Checks

Julian 21:31

Just a quick break. Let me ask you a question. How much of your last pull request did you actually write? And how much did AI write? If Copilot or ChatGPT disappeared tomorrow, would you still know how your code works and could you explain it in a code review? This is the problem we hear about the most from developers like you who reach out to us for a chat. Pie bytes, how PieBytes developer mindset program helps you become the developer who uses AI effectively, not the one who is completely proper. Buy it through a one-to-one coaching, real-world project, and proper code reviews with an expert code, not AI, you'll actually understand the Python code that you ship. If you're tired of feeling like a product engineer instead of a real developer, check out and apply for PDM using the link in the description. Now back to the episode.

Bob 22:27

Yeah, um, I would like to learn a bit more about the natural bias and how you solved or addressed it. Um yeah, can you talk a bit about that?

Aaron 22:38

Sure. Uh so we use a combination of of tools and and methodologies and and so on. So we we have a pipeline, a natural natural language processing pipeline that we use. We we connect it also with we're using Claude, anthropics claude models, um, and with Langchain, and we take all these kind of we extract all these multidimensional signals that we're getting in from the transcript. So when and and you know the audio recording, and we take these, you know, multidimensional kind of signals across various things like technical skills, behavioral traits, cultural alignment, um, and we we kind of take all of these and we structure them in a way that and we store them in our Django or M database model, and and you know, we take all of these and we go through a very kind of structured analysis that's um that's then checks itself. So we have a bunch of things. So we have red flag analysis, we have consistency analysis, we have when it comes to scoring, we have various uh levels of confidence, and we actually show that to the employers and the customers because we don't want it to come across that you know, this is how we scored it, but why? And you know, we still want the human in the loop to make the decision. So we give things like you know, our score, our confidence level to that score, but also the reasoning behind that, uh, so that they can make their own decisions on yes, I agree with that score, yeah, no, I don't. And then also we we I wrote in various um kind of checks and balances. So there's for every score, there's actually three different models that are competing with each other to agree. So what happens if you were to send one uh you know one model doing the analysis and it's you know weighing against the NLP stack, it's weighing against all of these other these weights that we have for the scoring, it could do a great job, but it could also you know not nail it exactly. So there's actually uh almost like a a I don't know how you want to say a table of experts, a panel of experts that are actually checking each other's responses. So they'll they'll all score, then they will review each other's score, and then they will all agree on the best combination and the best final score. So it's it's kind of it keeps itself in check and balance essentially.

Bob 25:18

Interesting. So yeah, like um is this then also going back to the AI unpredictability and and the hallucination we spoke about before. Um, that can happen, but then you have independent agents or models that then fact check it, so to say, and then it get flagged that maybe the original model then has to do the thing again, right? So something like that?

Learning By Building And Tenacity

Aaron 25:41

Yeah, something like that. And for every every analysis, we have a separate agent uh written within Langchain. And so essentially they're the prompting and the structure and all of the scoring is very specific. So it's not a generic send a transcript to an LLM and say score this. It's we have multiple agents that and every agent you know is very, very finely instructed. We have examples on you know what good, what bad looks like, and and and so we have, I think at this point we have 13 different agents that all all have a different task. And then when the actual AI avatar that we do is for the interaction, we provide very specific um prompting and guidelines and guardrails. And we also have multiple AI avatar agents that handle different parts of the interview process. So, you know, you'll have one that that will go for like say the first five minutes that's the intro, it does all that stuff, and then it hands it off in the background to the next one for a more technical part of the interview. The candidate doesn't notice the handoff, there's no visual uh you know indication that something's happening. But what we've done been able to, you know, identify is that by instead of having one generic form to do everything, having very small experts in each field that communicate with each other has dramatically increased the efficiency of and the quality of the scoring that we get out of it.

Bob 27:14

Cool, cool. So you definitely uh became an AI expert, right? Like, how did you learn this stuff? Uh, because did this you did it by yourself, or do you have a team of developers?

Aaron 27:24

Uh yeah, I just did it all by myself. Uh I've I'm very uh tenacious. I uh you know, when I want to learn something, I'll learn something. So it's just a lot of playing with it. A lot of I mean, I I I leverage, you know, coding assistance for, you know, like I at the very early stages, I would I'd have a bug and I'd throw it in the chat GPT and I'd be like, what does this mean? And then it tells me, and then I have to figure out what it's talking about, and then translate it into code. Now we have stuff like you know, cursor and clawed code that that can benefit. Um you still should check the code and understand what it's writing because you're the one gonna maintain it. So no, no one was spitting out. But you know, in the very early stages, it was you know, it was just googling and and copy and pasting and and reading this, and like it was a lot of getting there. And I think that that the idea that you know you're gonna be an expert right away or you know everything is is false. There's there's a world of information out there, and if you have the passion, uh, and in my case, I had an actual application I was trying to build, you know, I I funnel my passion into that and learning just came. Yeah.

Bob 28:38

Yeah. And this is not the first product you're building, right? Because when you joined PDM, uh you started working on this uh video encoding stuff, right? And that actually you built also out into a business, right? So this was not the first time.

Mindset: From Scripts To SaaS

Aaron 28:51

That's right. Yeah. So I did build out uh that platform, a much harder platform to put out, to be honest. I remember I remember working with you trying to write those lambda functions, and you know, there are some some times I was and you're before AI, 2020. Yeah, before AI, and I was just going through Google and you know, Reddit threads, and uh you know, but you learn, you learn, and I think that's the key is you know, you don't you're not gonna know everything and nobody's an expert in everything, but the learning how to learn and just having the tenacity to keep going at it is is critical.

Bob 29:31

Yeah, I feel um there's a lot of mindset involved, right? So do you have some tips for for people that are are stuck or they can write scripts and all that, but getting really to a potential SaaS product. Uh I mean you already we can already summarize some of the tips, right? Like willingness to try, willingness to fail, um doing a lot of trial and error. Um, but maybe there are some other practical, timeless things, even from your PDM days or from the the previous product that you just basically you're just just your hacking framework. Take the work every day.

Breakthroughs After Roadblocks

Aaron 30:11

Yeah, I would say I'd say from going from just writing scripts to writing a product is a few key things. I think for me, the critical part was uh getting over the imposter syndrome, you know. That you know, I wrote a small script, but this is basic and this is all I can do um to thinking bigger, you know. And I think when I came to these products, like you know, the the video encoder or now to job hive, it was what what's the big picture and and know kind of where you want it to go, but start small. I think the biggest issue a lot of people is they get excited about something or they want to build this big thing, but they burn out because it be feels too overwhelming. And that's I think that's the biggest problem is you you just wrote a script, a script is already got already one of the steps on the path you want to go. So then the next time write a little bit more of your script, write a little bit more of your script. It's just small incremental steps. And you know, same with with bodybuilding, you don't get you know huge in a day. It's the consistency in small. So I would say anyone who's even just writing some scripts now is you're on your first step. Alrighty, and and just keep going, just start doing nowhere you want to go, but work, you know, day by day, progressively. Just sit down and say, today I'm gonna add a little bit more to it, I'm gonna do a little bit more to it, do a little more to it. And then by the time you lift your head in a month, you'll be like, Wow, I'm already this far ahead. Um, so I I would think that, and then also you know, not give up because most of the time when you're so frustrated, and I've had these nights, and I'm sure you've had many of them, where you just are you just can't get past this bug, you're just so annoyed. This happened in PDM. I remember I think I even complained to you a bunch of times, but I couldn't get my head around it, and it was not moving forward, and it was, and so I just was like, Well, what's the point? Let's just scrap this and start something new. But it's always right then is when you're on the verge of a breakthrough, you know. So I would say just take a breath, go walk, take a walk, or do something, come back, and you'd be surprised, and then you make that breakthrough, and then all of that pain and all those little baby steps before are all worth it, and now you got you know, you're propelled to the next step.

Bob 32:33

So yeah, I remember those times. Um, if people watch this on YouTube, it's like the PDF was a bit like going up, and then there was like this dip, and then oh my god, maybe I should leave the program and boom spike, da-da-da-da. Gradual little dip. Oh my god, this is so hard. Boom. Right? It's it's like it's I I like that. Like, if if if you're completely stuck and you want to give up, you're actually on the brink on of some sort of breakthrough, right? You just don't see it in the moment, and maybe back looking back, oh that that made sense. Um yeah, so do they have an example with uh job hive or the the other uh app uh of one of those boweries that like this this no, I'm done. Like this is not going to uh go to MVP, you know? Any story from the trenches?

Aaron 33:21

Uh for job hive, yep. I've had many of those days. I've had I had a time where I was doing great. We got the odd the you know, the avatar is working properly, and and you know, um I feel like we're just smooth and sailing. And then I don't know what happened, but suddenly the the the AI avatar just wouldn't work, just decided it didn't want to talk to me anymore. And I went, I tore all the code apart. I you know, I was like, this is the most important part of the interview. You need to have a human, you need to have this interaction. If it I can't get it to work, what's the point? And it took me well about five days to figure it out. And you know, and then when I when I figured it out, and it was just something very small and nested that I wasn't paying attention to, it was it suddenly because what I was trying to improve it, and so I somewhere along the line, am I improving it? I I messed it up a little bit. Uh, but then once I figured out that issue, now it kind of just worked, and what I was trying to make the improvement worked, and it's like I just suddenly in this new realm and new stage, and now I'm excited again and I keep going. But for those five days of of a broken avatar not talking to me, I was like, What's the point anymore?

Bob 34:51

Yeah, yeah. If your core logic is not working, it's like your private platform and the lambda is not working, so you cannot submit your code. And it was some sort of regression.

Aaron 35:04

But yeah, yeah, yeah, yeah. It was uh I can't what was enough, I can't even remember, but it was I can't even remember at this point. I think I my mind blocked it out, but it was five days, right? Yeah, it was small enough that it it didn't it didn't break everything else, just that, but it was big enough that it blocked, you know, so much.

Beta Launch And How To Try JobHive

Bob 35:27

So yeah, and then I think because you're using so many um external dependencies and systems, it's it's harder to debug, right? Because um it's it's not only your code, you hand it off, you call an API, and there's now more yeah, combinations and and well dependencies, right? And that makes it sometimes harder to uh to debug. Exactly, yeah. Yeah, yeah. Nice. Cool. Um yeah, anything else you want to share about uh job hive, uh as we said, right? Launching phase. Uh can people just try it out from the website or wherever should they?

Aaron 36:02

Yeah, so we we are we're currently having a very small beta with uh onboarding a few cut companies to to test it out. If anyone is willing to test it out from an employer point of view, you can create an employer account or a candidate point of view to test it out uh and provide feedback, that'd be fantastic. So, yeah, uh you can reach out to me or Bob can pass it along and uh yeah, I'd love to get some more feedback.

Bob 36:30

Yeah, we'll link it. I will link your details as well if they want to reach out. Yeah, yeah. No, cool. Uh yeah, I think the backstory was way more interesting than a UI overview. I mean, we could still do that, but people can just also just go and and check it out, and there are probably like demo videos there. So yeah, amazing, cool. Um, yeah, what what else influenced uh you in your coding? You mentioned working out and the grid and persistence. That has that been an influence on uh making this work?

Fitness, Discipline, And Clear Thinking

Aaron 37:01

I definitely think so. I've always been uh for many years now have you know uh quite a passion for bodybuilding. My wife also did and won a bunch of competitions. I I never went that far, but but uh yeah, I found I found that for me uh that the gym was a great extension for coding in that it allows me to have a break when I'm frustrated. But I've also noticed that being active, getting up, you know, working out has allowed me to to think things sharper and clearer. So I think being active in the gem for me has actually helped me in many areas, including coding, uh, you know, to be able to focus and to have a you know a clearer mindset.

Bob 37:50

Yeah, because if you're talking about bodybuilding, that's that's pretty disciplined, right? Because that's not only the workout and and not just doing your workout, but going progressively harder, and then the diet that comes with it, right? That's even if you're not competing, it can be uh it can still be pretty uh disciplined thing, right? Yes, it can.

Aaron 38:09

Yeah.

Bob 38:10

Nice. Well, we always uh wrap up with uh what you're reading and and book tips. So uh what are you reading? Uh and or do you have a book tip?

Book Pick: Pitch Anything

Aaron 38:19

Uh I'm currently reading a bunch of blogs on platform engineering, but I would say a recommendation, actually, one that I picked up again not too long ago, was a book called Pitch Anything uh by Orrin Claff. Not super, super well known, but a very fascinating book and oddly applicable to almost every area of your life. So, you know, if you're not familiar with it, he's uh you know, he brings in investments, he does pitches and raises money. So the whole book is talking about how to present yourself, how to communicate. And and you know that that when you're starting with something and communicating with someone, there's often what they call the croc brain. So the person receiving the information is naturally, you know, you know, defensive and to what you're saying. Someone just talks about breaking down the croc brain and how to present yourself and communicate. And the reason I like that book and why I recommend it is in pretty much any area of your life, at some point, and even on a day-to-day basis, you're communicating your thoughts, you're communicating to another person. And in this case, anyone listening, if you're wanting to build a product and launch it, you're at some point gonna have to communicate what your product does, or talk to investors, or talk to customers. And being able to communicate your vision uh in in an articulate way is super critical. And and this book is is a great way, you know, you can take these principles and apply them. So I would recommend Pitch Anything but or glove. And uh yeah.

Closing Thoughts And Community Invite

Bob 39:50

That's it's really funny. I did you recommend that book. I was listening to a podcast today, and I think it was on founders, whatever, and they also uh recommended that book. So uh I I now need to buy it like straight because that's like two recommendations one day. But yeah, that's a great point, right? Like uh, and again, we as developers we focus so much on the code and then the technical part, and that's kind of what gets us uh if you're in the building phase, easily to 60 hours a week. But then you still have to well, you might not be the CEO, right, at the same time, but plus probably with a startup situation, you are at least a CTO or you have influence in that process, and then you communication skills are critical, right? So it can be a brilliant solution. But if you don't know how to pitch it, right, then we're still nowhere. So uh we have to wear all these hats. Yeah. Yeah. Nice. Well, thanks for uh sharing your story. Uh very interesting. I'm sure people will enjoy it and uh hope they will check it out, reach out to you. Uh any final shout out or recommendations, piece of advice, something you want to share before we wrap up?

Aaron 41:00

Uh I would say just keep following PDM. Yeah, they got some great stuff. I love watching what you guys are doing. And uh yeah, just keep plugging away, like I say, if you got an idea or you're working on something, small baby steps, and you'll get there.

Bob 41:12

Awesome. Well, thanks a lot, Aaron. And uh yeah, I wish you uh good luck with uh with this uh startup.

Aaron 41:21

Awesome. Thanks so much, Bob.

Bob 41:22

Thanks for joining. All right, see you back.

Julian 41:25

Hey everyone, thanks for tuning into the Pie Bytes podcast. I really hope you enjoyed it. A quick message from me and Bob before you go to get the most out of your experience with Pie Bytes, including learning more Python, engaging with other developers, learning about our guests, discussing these podcast episodes, and much, much more. Please join our community at pybytes.circle.so. The link is on the screen if you're watching this on YouTube, and it's in the show notes for everyone else. When you join, make sure you introduce yourself, engage with myself and Bob, and the many other developers in the community. It's one of the greatest things you can do to expand your knowledge and reach and network as a Python developer. We'll see you in the next episode, and we will see you in the community.