How Hackathons Make You a Better Software Engineer

Today I'm joined by Perus Pulatrac, Software Engineer over at booking.com and apparently he's a living legend. If you ask anyone about hackathons within Booking, they'll point you to him. He's actually won a couple or so, and with close to 1000 engineers participating, that's quite the accomplishment. It's one of his main passions and he's going to explain why hackathons are valuable for you as software engineers and for a company to organize.

They even organize hackathons in their graduate programs and there's a lot of skills you get from participating. After this conversation, I'm sure you'll want to participate as well. So enjoy. Do you do that often? Go from an idea and kind of a need where you see there's no real good solution. Or even when it's just an experiment, you create a tool out of that. This is kind of my hobby, So

what I'm doing is like each time I have some sort of idea or problem in my life, it doesn't matter what it is, I'm always looking how can I like solve this problem for myself and also anyone has the similar problem. So I did it for multiple applications. Some of them are already published, some of them I, I didn't publish yet and like it. It could be like, for example, I created something for appointment Hunter and this is what this is like.

Getting appointment for some government was really challenging for me and I've seen people struggling and you need to go to the page, refresh and refresh and so on. So I said why we spending a lot of time doing that? We can just automate it. So I did it in like even less than a less than an hour. I create an application, it goes and does everything for you, sending notification when it's done. So you can do the same thing. You can enable yourself for the job of work that is like you're

struggling now. And I shared this application and apparently that was super interesting for the HR of the company I was working and they share it with all people and they were really happy. I got a lot of feedback, but that was like hundreds line of JavaScript code. So there's nothing really challenging, but I could do it like in an. So this is my hobby to be

honest. I even develop games, I develop some application for myself, develop even for different technology, not not for the web itself, like using Electron, I develop application for Macintosh and like other OS. And that's really feasible. Like if you know the ground knowledge, you can easily, there's no need to go to documentation, right? Like before and go and read a lot of documentation, understand all aspects. You need to know the general capabilities of technologies.

Then you can easily puzzle that up and make what you want.

I feel like building and creating applications regardless of the platform, even seeing problems and then trying to find a solution in in software, it's a real skill. And I'm wondering from your perspective, since you say this is a hobby of mine, how do you become good at that? I think the best thing to being a good at this skill is doing it. So it's like you need to try, you need to fail, you need to like do it more and more.

And I always like, I remember the first, first like application I was started, I struggled a lot. It was not really working well. But I even started this hobby before this AI era. So that was more like challenging and time consuming for me from the start of this era. It was like a really happy time for me to be honest because I jumped into the project in a day. So I even created one like optimization application for myself. I have these printers that you have in the stores, thermal,

thermal printers. And for my daily work, it automatically prints out my to do tasks. So I created with JavaScript connecting with some server and web hook and just that's patched everything like in a couple hours and I have like a printer everyday prints out my To Do List. This is like really like like old fashioned way of doing that. But I want to just experiment doing different stuff, making some crazy and that like this AI tools enabled me to do it super fast.

So I did it like so many things like this before, but it took me a month or two months sometimes to and like your vibe is not always you cannot do it really fast. But now when I start, most of the time I finish the same day. So it, it doesn't like take more than one day for me to getting something. The production ready stayed. But the code, most of the code is right only. So it's not that high quality so but it works with some because I'm not going to sell those product.

But this is like MVP. But it it enable me and resolve my problem, I can go ahead. I also tried to make one of the projects in high quality, so that took more time, but I could break it down with the help of AI with the standard I know like each function should be really short, like file name, everything. I could also do it the same way, but it takes way more time and you should start it really soon.

You should not like make the product working then switching to change it. So it's really horrible. You need to start from the ground, don't allow the AI go the wrong way. Pluck it and go step by step. It takes more time but the end result can be more production ready. If you want to sell it, if you want to have something that that is maintainable later.

But this product, most of the product I generate is not really maintainable, but it does the job I want and that's enough for me. There's no like security really layer in some of them and some of them which need security, I have even four or five layer of security. Because if if you know it and you know, you see this is like a challenge for the security can already ask AI to do that or you need to do it manually. Sometimes it's not always good at doing everything to get the security.

Yeah, that makes sense. I mean, this the skill of getting something up and running in one day or even a few days, it really reminds me of hackathons where there's a competition and this is the goal. Basically everyone comes together either as individuals or as groups that just have formed that meet this goal that they have set, and then they have a few days to execute on it. What has been your experience

with hackathons? Exactly, I'm passionate about hackathons and in my company I'm like a symbol of hackathon. You're the man. And yeah, like every time speaking with people, they start talking about hackathons. I even like the one that going out from the like representative of the company going out and speaking about hackathon universities as well, because this is like really my really

real passion. So that's the same same reason because this comes from my hobby that is generating something and like ideating, solving the problem and that goes to the hackathon as well. And I found like the opportunity in my company also to having these availability to having 2 *

a year hackathons. And I've done a lot of crazy project in hackathons, which I'm really proud of because the end result was not only in the hackathon, I could like get something out to the world and like get real result out of that. And that was really interesting. I didn't know that it's really even possible in the past that you can do something hackathon and at the end you get something

real life out there. You have an example of what you built in the hackathon and what made it in the product.

Exactly, I have quite a few example. I can start with the one of them, which was the start of the ChatGPT era. We, we formed A-Team to create the agent. So our mindset was to having something like before that we had some people, you go there and just they manage everything for, for your trip. I wanted to create the same thing with the AI. That was really like crazy idea at that time because our company

like booking.com is huge. So it's you cannot easily attach everything to the like simple agent. So we started with the attraction only. So to look into the attraction, we created the hackathon and we started working on that. And what we did was to create an agent that you can chat and it owes already encrypted with the generative UI. And it could generate UI for the list and show you show you a list of item.

Then you can buy or select, add to your like selections or save it as like a bookmark and so on. Then you can ask question about your whole trip. So for example, say you are in Paris and you want to do all activities, restaurant and some of them even not available at booking.com but you want to plan it, you want to have itinerary for yourself. So the idea was to discuss, ask questions.

So you can say I have like with two children, a dog and like these circumstances, it can be even complex circumstances and it's not even easy for human or everyone to manage that. But AI can easily break down. And also we created something like vector database to search through all availability that we have and serve better result for people. It was just for the attraction. And that was quite nice experience for me that we developed something crazy.

There was even no, that time I remember Open AI, there was like API for open AI, but the ChatGPT quality was different between different with the API that we had available globally. So what we did was to creating a wrapper over the ChatGPT and like a hacking, hacking way of getting access to ChatGPT because the quality was better. We wanted a really high quality, high quality LLM to analyse everything that we had because hallucination was really a thing that time and now it's way

better than before. Then we even learn prompt engineering throughout the process. So the, the, IT was like a lot of learning in in a week and that's crazy learning and applying. And the other end of the day, we also won't. It was really, really nice feeling as well. So also encouraged me, OK, I should like to spend more time focusing and looking for different ideas. And the main important thing I also mentioned is this is not ending when you won't or finish the process.

It's just a start. Like the company has started a huge team over the idea and many other people started to develop the same thing, but with the standard approach. Because the Hackerton code is not reusable, it's right only again. And they started to form a team and develop and now we have AI3 planner already developed and available in many regions. And I'm super proud that I like started this fire and now it's like helping a lot of people.

And this is really, really great feeling that I also had with my small project, like personally, now I can do it in a larger scale. I'm super happy and really, really nice. That's really amazing. I mean only the fact of booking has thousands of engineers and if they all get twice a year time to execute on projects like this and the really good ideas get follow through and actually make it in the product, that is such great fulfilment.

It's incredible. Exactly, I last time I can mention it was it was like around 900 people involved. Yeah, in the hackathon. And that was a lot. If you walk into the office, see the why people like drawing something on the whiteboard, like discussing. And this is not only about like one role in to hackathon. You can easily switch roles and you can experiment.

For example, if you like to be engineering manager and you are engineer in the hackathon there there's opportunity to go to the role and practice and see if you are really good at it. Yeah, or data engineering, any other role, anything that's not what you're doing. This is a really good moment to go to the like different role and see if you like it or not. So this is really also a really good point for the hackathon. So it's not just engineering that participates.

Exactly 1 of the project I can mention that was like around we were around 12 people in the team. That was a really huge team for that project. And we had SEO engineer, data analysis, MLPM, EM, all different roles even copywriter. First I was like thinking, is that even needed for like what? Are they going to do?

What they're going to do. But at the end of the day, I've seen each of them has a crucial like point about everything we do. And we, we have really nice, like, copywriter giving us right wording for the advertisement, for the landing pages, for whatever we create. And the quality at last was like killing quality. So it was like no brainer for the team to say, OK, you are the

winner. Yeah. And that that time was I realized what is the value of these like rows and I learned a lot about different rows, what is like their mindset. I didn't know about SEO, how they're going through these steps, how did they find keywords and other, other stuff around it. But this is a great learning if you are, if you're not really happy with what you're doing, This is also a good moment to see other opportunities, what there are people doing.

If you are interested in you can jump into the different role and you can transition also as well. Yeah, I really like how you make the most out of the opportunity, not just from an idea sense, but from a learning perspective. Learning how to collaborate with people that you otherwise will never speak kind of in your day-to-day and seeing and kind of gaining empathy for their perspective and what they're greater and see how that fits in a product. I only see upsides basically.

I also participated in a hackathon, Xavier. We get a one day every other month, which is great, but it's only one day. And then usually you do like a proof of concept, you try and do as much in one day. And then my previous experience at a previous client, it was like a 2 1/2 day thing. We formed a group and I was like, OK, this is going to be fun because I never get more than one day. So we can actually do a lot more

now. And I had this one person who's team captain, he was like, I've done many hackathons, I've done like 40. OK, we're going to decide idea we're going to have he had things time framed. He really was the one that kept everything on track, make sure people made good trade-offs, were practical, also delivered and executed and didn't just overthink or overanalog. And I think that is really what I needed. Exactly, so I wasn't that person, but I tend to be that

person. So now I am the person that you mentioned. So I'm when I the first day, I always like gather all team. We need to have a strategy. We need to look into the people that we are like presenting to. So even list of people that are like in the jury, what is their mindset? What we should target everything. So this is like selling the real product out there. So if you want to sell your product, you need to look at the customer who is going to like

assess my product. You need to like or like engineering the process and not just OK, my idea is good, I develop it. You need to do it like like to the point and develop something and especially for the demo, like demo is a really real crucial part of the hackathon and you need to nail it. And she's not like you have a great product, sit it there and just your presentation is tool page and you cannot represent what you've done.

What is the like crucial value behind it For one of the hackathon, we even created like a production ready videos like even we interviewed with some people out outside of the company. So if we like attached a lot of like content and to make it tangible and also we look into the feeling of people when you create something, you need to also touch their feeling. It's not like, OK, my product does this good. You need to touch the feeling.

So you need to go through the all aspect of a real company that when you're selling a small start up, this is exactly the same as selling 1 hackathon project. You need to go through the same steps and just think like that. Then it changes everything and all people are getting energy. OK, we have it like mindset now we can go to this and then the brainstorming coming and people like sharing their expertise. We had some people OK, I have a a really good expertise in

narration. I can narrow like create a high quality narration for the videos. So what we we created at the end of the day even was amazed of the quality because so OK, I am really good at editing video. So they were like like PMEMS but they had some expertise. They can just turn it into the project and make really the most

out of the project. I think that's gold, like treating your hackathon project like a mini startup, seeing what you have in people and leveraging whatever creativity they have to execute and tell a story and not just create software, but also sell it to whomever is judging you or whomever the audience is. Exactly.

That's fantastic. That makes it a lot more like I get super excited hearing you talk about that environment. It's really cool that you get time from work to do that. Have you also seen that outside of company settings like hackathons that are organized by communities that have the same vigor? I can say 1 is still is still inside but it's like we have for the graduates. So we have same setup for graduate, but it is not one

week, it's 4 weeks. So it is even you have even more time and that is super amazing because I'm a mentor of those program. I'm aware of all the steps that happens. So past year it was like 4. You have 4 weeks, you join as a graduate, you don't know anything about like working at the company. You just jump out of the university. We want you to know everything after months. So this is a hackatone you when you join to the first day, you enter to one of the idea. So it's like joining to

hackatone. You form a team and start working on the on your project. And each week we do, we'll have some specific goals. For example, first week you learn agile, you learn how to work and collaborate with the teams. Second week you do front end or back end and 3rd, 3rd week you do front end. And last week you just form it, create demo and present it. So these like step by step process. It's like a month that you have, but it's super intense. You need to learn a lot of things.

Yeah, imagine you don't know Java, you need to learn it and you need to do it as well. So with the other help of AI, it's a bit faster and easier. But in something we do not even allow them to do, do AI to to try and struggle and know the value of the stuff, understand and know the technology. What is the right way of doing that? Then later this tool can enable you to do that like faster and so on. But that was really amazing opportunity for people. And I was mentor of 1 mentor of

one team. I've seen the first day they struggle even to speak together, but last day of the week, after like learning sessions and process that we had, they were really close to each other. They like challenge each other, they discuss really critical stuff with each other and that this is exactly what we want as a company. After a month, if you join to a team, you have almost no

onboarding. You already know all technology at the company, you already enable yourself using everything that you probably use later in your team. And this is going to reduce the onboarding process. And the feedback we got from the engineer managers was amazing. They were really amazed with people that Giant just joined from the university with that level of understanding of the teamwork and like collaboration with the other people.

And that was one of the example I can say some sort of external which was really nice. Yeah, interesting. This for me is fascinating because I've also seen the industry trend of we don't need junior engineers anymore.

We have AI that does that, even automate it. And then our seniors, they review PRS and stuff like that. But the fact that you have junior programs and graduate programs and people come in, they execute for four weeks and then they have real praise from engineering managers on that level is fantastic. Exactly, I'm completely against this idea because what I've seen as a result, like I was working directly with them, I was like mentoring them daily.

What I've seen they are really great and we need it like we need juniors because they, they come from like university, they have fresh mind. We even had juniors that coming from university and they had like measured in LLM. So they have like top knowledge about LLM. We want them because if they we move them to to the right team, they can like perform 10 times better than other people. They do not know, do not know anything about LLM, for example.

And throughout this process we could like enable them in a month. So I don't think it's still like they're not valuable. They're super valuable and they could like perform really well. So I know a lot of them. I'm now friend with them. I wasn't their mental, but they're really performed really well. And all people even like engineers are amazed. Is that really a graduate that joined our team?

They're really good and they can perform really well and with the help of AI, if they're using it the right way and first go through the understanding of the factors, not jumping into the conclusion, they always can have a proper result as well in consultation with the other seniors around them. Yeah, sounds like a lot of fun. Like if I could go back in time, I would have loved to join a graduate program. Like this? Definitely.

Sounds like so much fun. Exactly how many people are usually in the graduate program that do this for a week? Last time was like 770. 70 Yeah, that's incredible, yeah. Really huge amount of people and they're really crazy with the idea they have a fresh mind. They come up with their own

ideas. There's some sort of like ground ideas coming from the product managers because you don't want to just like put their like what they do just they want to apply something into it and they want to use those projects and they want we want to treat it as a like production product to make them also like doing something serious, not just OK, I'm going to do it and they're going to just delete it. At the end of the day, this is not the case.

You want to make them more serious like this is something that we're going to use it later. So they are mostly OK. The algorithm should be like this. And so the discussion changes when you note the thing that your looking is seriously going to be used later. But they have like capability to change it to to what extent they want because they can add a new feature, they can change the direction, but the main goal is

like already defined. You want to have something with XYZ, but you can do whatever you want. So there's no limit into the functionality and features. They can just come up with the crazy ideas if they can implement it. I feel like that is very smart and I don't know if I've heard that before.

Like if I think of hackathons, it's usually you have a theme and you can do whatever you want within that theme, but if you are creating something and there's a chance that it will be used in the eventual product, you act differently. I think that's very smart, especially for a product company to say, OK, these are our products and product managers have thought of these ideas. You can bring in your own or you can participate in this and then it might actually contribute and

benefit. Usually when you're talking about hackathons, if you talk to the wrong people, they see it as a cost, right? What do you mean 3 days? People don't work and they just play around what? What is the benefit? Basically, you can say something vague with regards to knowledge sharing and learning, stuff like that, but if in the end there's something tangible that comes out of it that we can actually put to production and has business impact. Exactly.

And all of a sudden you're like, oh, that's interesting. Yeah, what what like what the booking does is like goes from the other side, it's sending a message to all PMS. Do you have anything you need and you do not have capability of doing that. So they're listing those up and then they're looking into the items. OK, which one is more suitable for these people that are coming because the project itself built with the like mentorship. It's not like they're developing by themselves.

There's a mentor always looking into the all process. So the quality, the the output is really high, the quality of the output and they cannot be used directly, but they are really ground like floor for developing a real like production ready product out. So we're using high tech, the high technology and like cloud and everything that you're using daily because you want them to experience what they're they're

trying to do later. So they're like small amount of experience before jumping into real work and you experience everything. Like we try to allow them to experience all aspects because in large, large organizations, you may go to a team that you have like working only a specific on a specific feature. So you do not allow to work on everything. But we want them to already

experience everything. Different aspects of the like software engineering and some of them even rotate into different teams, like 6-1 to the team one and six one to team 2. Because if we want to want them to have exposure to different aspect of the product because it's too big and at least a lot of stuff to learn. And this is like a really ground learning process. And then they jump to the like 2-6 months process that they, they will rotate into different

teams and learn more and more. And after that they they get automatic promotion and then they'll go to the real life team and stop like stay in one team. And this process I think is really a smart design, has a really smart design that you like stop drum intense when the first day that you work at the team, you feel like, OK, this is a relief for me.

So, but they learned a lot. So the process, they're learning a lot and they learn not only for hard skill, but we had huge focus on the soft skill and agile and like teamwork. So we have so many more programming, like per programming, so, so many things that to connect them with with each other and know how to work with each other in real, real life. Yeah. And that's really important. Yeah, interesting.

And you've been in many hackathon teams and you've mentored graduates that also go through this process. What are some of the things that people struggle with with regards to executing and delivering and kind of reaching the goal that they set for themselves? Is it planning? Is it actually teamwork or execution? What is usually a challenge?

One of the challenge I've seen is when, when you open like the capability of the new features, they may like aim for the really high and they cannot reach it. So they, they do not know how to easily break down the challenges to the small challenges and go step by step. Sometimes they aim really high and they cannot even reach it. So they're disappointed, they change direction and so on. So as a mentor, what I do is always sit first, sit and look. They're struggling.

So I do not like jump in. You need to wait and see or how they're struggling. What is their then you need to guide them, not even directly give them the answer. And what if we do this And what if we do that? So this way you help them in order to think that way and go ahead and figure out the solution by themselves. So I, I, when I just guide them through the, these challenges, they end. OK, We should do it like small

by step by step. We should break it down if and then they managed to do it. And that was one of the thing I've seen. And the other thing was more about the collaboration, the the issue that he had, they had they didn't know how to collaborate well with each other. Everyone was really great solo, but they couldn't work together. They couldn't enable each other and that was also something that we were working a lot on on the process to enable them.

We did like more programming is one of the thing that starts and breaks the ice between them. Yeah, because one of them you need to see it and the other need to guide. And so did this speaking happens technical speaking question answer. So the desk, this is really nice start world for the for the enabling them for the collaboration. Yeah. Is that the same for experienced engineers?

I, I can imagine that if I've never done a hackathon and I go into a team, I do kind of know how to break stuff down. I might fall into kind of analysis paralysis. Collaboration might always be challenging depending on if you've more programmed, if you know how to break those barriers. But it's are those the same challenges you've seen in existing hackathon teams? I would say kind of kind of the same challenges, but each hackathon team needs someone to guide and lead.

So that's why I always encourage EMS to join the hackathons because they already have the capability and knowledge of managing teams. And if they join hackathons, most of the time they're not joining because, OK, it's too technical, I don't want to join it. But we want EMS in hackathons and if they join, they can easily guide and help and do whatever they do in real life, but a bit more fast in a bit more fast peace environment.

They can guide people, they can like lead people in order to each team. A successful team hackathon needs someone to do that, even if it's not like he or she is not a an engineer manager. It needs someone to manage everything and blocks ideas because so many ideas going to come up and we need to like have a some limit for our focus on what we're going to deliver and then expand, expand like when times like times time allows. So that's really important.

Like member of each successful hackathon teams. If all like engineers sitting together just coding even the quality, if the quality of the product is super great. They do do like ground breaking project is there, it's not going to be recognized. Yeah. And they need someone to focus on the goal and like why we doing that always should be this question and then getting to the result then adding up some like fancy feature over that.

I receive a lot of questions for people that are slightly more early in career with regards to how can I distinguish myself. Nowadays in this tech field, people are getting laid off, there's very good expertise out in the market and I'm applying for jobs, but how can I distinguish myself? I feel like there's a lot of benefits that you can get from hackathon, but I'm very curious to hear your perspective. What would you say the biggest benefits are of participating in

in hackathons as engineers? For me, for for me myself, it was like the most benefit I get was not like the prize was not the, the like other stuff around.

It was the learning process I had. So for the softer skill, you have huge amount of opportunity to to learn because you have one week. You need to connect with people, you need to collaborate with people, you need to get a like something out with those people. So this is really challenging. We normally when you like on boarding to a team, it takes a month or something to get the same like connection with the people.

But because you, you have this urge, you don't have time, it allows you to jump and it allows to over perform and doing that. So, so it was like for me, it was really nice. It's not even possible to connect for those people. And now they're like all teams. I was working now they're my

close friends. So it was like the time we were really in connection that we changed to be a friend because even like more than like my connection is more with like my day-to-day, like teammates because time is really intense. You need to like focus. You need to put your whole attention into the project and then you need to collaborate well with all people. And not only for that, for the

technical aspect is amazing. For example, for prompt engineering, that was like my enter point to the prompt engineering understanding about that, like getting my hand dirty with with AILM. So maybe your day-to-day work, you don't have capability to do that. So you're doing working something completely not related to that. But hackathon is a great time to

to just go out there. Even if you find a project that is that they're using some technology that you like, just jump into it. This is the best moment you can learn. You can apply the knowledge in real life project and I'm using it always like like this because I'm also like into gaming and developing games, but really hobby games. As I mentioned, I found same like opportunity in one of the hackathon there like there was a team, they wanted to develop a game.

I jump into the idea because, OK, this is a really good moment. I cannot find any opportunity to develop a game, a booking. So it doesn't have any relation. So the game itself was was to like enable people to learn. So we have a an experimentation, famous experimentation tooling that we developed for a long time. And because like experimentation is DNA of booking.com. We have so many AB, TS and so on, But the tool getting really complex because it developed throughout many years.

And if you want to learn it, it takes a lot of time. So the idea was to create a game and let people play the game and you getting to to a blocker and then you need to open experimentation tool and do some action. Then you go go to next step and it asks you now unblock. So you need to you you learn a process without even opening

documentation. It's like a it was like a Star Wars style game and you need to go to the step by step answering the question and some sometimes open this link and go to the experimentation. Now you need to set up the experimentation and go ahead. And because of the time, the experimentation always time is important because you need to wait for a month to get the result. But we speed that up with create like simulating the date and

data. We inject data to the experimentation like simulating that like a month past. Now you need to make decision. Now you need to make a call. Is that going to be full on or not? So this is really crucial and important decision for the company. But we changed it to a game and you can play thousand time and then you can understand and learn how to make the call in different situation. And the benefit goes directly to the company because this is

really crucial. If you call and make it run call, you're going to get like a lot lost a lot of money. So we created a game. You do not need to open even a documentation and go through the learning process super fast. So it has different levels. You start really simple experiment getting difficult, difficult and difficult. And that was only in a week. So it was really amazed. The team were really happy that we could manage this and the

output was really tangible. Maybe this is not for everyone. Some people are like video and so on, but there is a lot of people they want to get hand hands on the project and see what is what he's doing in real life and make a call and see if he's not. And it's like like safe call. So if you make this like wrong decision, you just learn and your company doesn't lose anything.

And that was really important. This is like one of the thing I mentioned that you need if you want to learn something, this is the best moment. You need to look into project that they have a goal or aim to do something that you want to learn. Jump into it.

There is no need to having the expertise to jump. This is the moment you need to go and learn and do it. Yeah, And I feel like what what would hold people back is they're afraid, right? How can I do something that I've never done before? I feel like nowadays that barrier of entry is lower and lower because you have a lot of tooling exactly and you are able to educate yourself. So then it's just confidence. It's way better than before. Yeah. Yeah, exactly.

So you just have to do it, man. I do hate that answer because for me it's very hard to start. And then people say, you know what makes it easier to start? It's just to do it and to do it over and over and over again. It sounds very simple and still it's challenging, right? Building something, having ideas. I mean, even when you and I talked over lunch just now, you were already talking about ideas and you see those ideas due to the experience you have building

up on your ideas. I feel like sometimes I have ideas and the step to actually building those, they do become smaller. I have fun with vibe coding. I know start easier, but I don't finish. For example, I have ideas that I'm like, oh, can I make a game out of that? Then I get to a point where I can make a lobby, people can join on the same room, but then I don't actually continue with creating the game. I start and do something else which I've never had before.

So the dynamic changes, but I am learning a lot. I feel like this pressure cooker, both for your soft skills and for your hard skills, is very valuable. Exactly the same, but about the idea is like I've created a project, it's like a personal project. I wanted to know more about sentiment analysis, like and like intention and topic detection and so on. But I wanted to do it in a safe environment for my personal data. Yeah. So I've created a Telegram bot and attach it to the friend

group that we have. So there's a group that we have our friends and this is a safe place. I can easily attach everything to that and this telegram bot just collect all messages and send it back to my LLM server that I have. And what I do, I analyse the like sentiment, the topics people speaking about is the trend goes up, goes down. I create a dashboard, fancy dashboard. But each time I'm looking at it, because this was a basic idea, I come up with a new idea.

OK, what if I predict the next message? So I added something to predict in my next message. So each time the message I get the context of the previous messages and message they reply and try to predict what is the next message. Then it compares with the next message. What is the accuracy of the

prediction? So there are many ideas coming up. Each time you look into the previous idea, you need to start the the, the the meaning of start is to if you have something there, it's like giving you energy to to add some new brick back over that. And then you see you ending huge project. But sometimes you aim really high and you want to, for example, add something. And this is really too big to be very feasible and to disappoint you to do it.

So need to go really small, really like much stuff, just go really small steps, small steps. Then then it encourages you to do it more and more. So this is for me it was working like doing really small features. So each time I didn't develop the whole idea together. The first version only had like showing the message that was in the tag of nothing out without LLM. But the current version, even I can switch between LLM. I can reanalyze. You can change the prompt.

So each time I added a new feature, I then it it gives me huge energy. It's adding a new feature to it. Let's do it. So even in midnight, I would say, OK, I need to add something to it. So I'd see what's going to happen. And this is like also sharing some of that with my friend and amaze. Oh my God, have you done that? So it's so nice and that's really nice. I think this stuff like you can do and the learning process, like I'm doing the same work at my current job nowadays, but the

scale is huge. This is completely different. But this learning process helped me really nice and like enabled me to do it involved in all steps, which was even not like part of my job description. I could involve. I could like have a huge and dramatic effect into many like prompting of the process and any other process related to LLM and AI related to my real job. But it started all from the hobby project. It's like my I'm kind of addicted to this nowadays. I can see that.

I think it's, I think it's very smart to figure out a way to get really good at executing with different technologies, building little projects here and there for your own kind of personal journey. I feel like it's really good in distinguishing yourself when you're talking about career perspective. But also like you mentioned, you can apply it at your current job. You can actually see other teams that are doing something that actually you've done in a personal project or you're like,

huh, how would you do that? You execute and then because you're having so much fun, you move to that team. I think it's fantastic. Yeah, exactly, exactly. And that helped me a lot. You know, for example, the same project I mentioned. I'm doing the same but with the larger scale and. Booking scale. In the booking scale, so, so I, I know both aspects. I know like how how it goes with the small project and how it goes. What's the challenges of the big

projects. So, and I can apply some of them to to the other. And even I learned something my job. OK, let's apply to my personal project. Yeah. And improve it. And so we had so, so many meetings with open AI that we are working directly. And I learned a lot about how, how you need to like, for example, use model. You need to solve with a larger model, then you need to adopt it to the medium model. And the final project you can use is be the flash, for

example. But you need to fine tune, you need to like adopt it. Don't start with the like cheaper model because you're going to get the poor result and you don't know how to like fine tune the the prompt that you have. So this learning a really general learning that you can use in your personal project or the real life project, but you can experiment it easily because if you want to experiment in the real world, it's more

challenging. But experimenting in my personal project related to my friend channel is super simple. And also I can experiment whatever I want, different languages, what was like the output for the different models and everything. And it goes with no risk. Yeah, I think that's the best. I I love learning about your projects and your passion for hackathon. Thank you so much for coming on and sharing. This was a blast. Thank you for having me. Cool. We'll round it off here.

Then, if you're still listening, leave a like. If you like the episode, let us know in the comments section what you thought and we'll see you on the next one.