PurePerformance - Mobile, AI, LLMs, Observability & Resiliency - Key Topics for Banks in Hungary with Adam Gajdi

Episode Date: December 18, 2023

I was invited to speak at BankTechShow in Budapest, Hungary where the nations IT leaders in the banking sector presented and discussed the future of banking - both in the cloud as well as what it means for the physical bank branches. I got a chance to sit down with Adam Gajdi, IT Solutions CoE Lead at K&H, who walked me through the process of their recent new mobile banking app launch. Adam highlighted the importance of observability for both business owners as well as developers. Furthermore, Adam enlightened me with the fact that Hungarian banks are mandated to conduct chaos tests to proof that their systems are resilient in case of data center outages. I was obviously also curious about how AI, LLMs and other technologies are adopted in their sector. Tune in to learn more

It's time for Pure Performance! Get your stopwatches ready, it's time at Pure Performance Café. I happen to be here in Budapest, in Hungary. I was invited yesterday to speak at a financial conference. It was called Bankotech. Unfortunately, my Hungarian is not really good, so I could only listen to a couple of conversations, but I have one of the

speakers with me today Adam hi hi nice to meet you nice to meet you too thanks for inviting me also to your office Adam could you do me a favor can you just quickly introduce yourself what you are and what you do sure basically I'm head of the solution of the digital channels here in the bank. So basically my team is responsible for the architectural parts of the mobile bank, of the e-bank and basically all the digital channels where customers meet our bank. And can you let me know yesterday, because I said my Hungarian is non-existent, so I did not know everything that was talked about what are the top topics right now that are impacting you in the financial industry here in Hungary yeah sure basically we heard a lot about cloud we heard a lot about AI and also some of the observability thing what we have have also presented during our presentation.

Yeah, maybe let's jump right into this because I know you were on stage and you talked about your observability when you had your latest mobile rollout. Can you tell me a little bit more about that process and why observability was so important for you? Yeah, sure. Basically, our new mobile app was released last summer, I mean this summer, this year. Basically it was a pretty exciting project for all of us because it's a brand new application, not an upgraded one.

And also it was released together with a brand new architecture, which is based on microservices. So it's a brand new application as a whole. and capturing all the details within this solution from the mobile application, from the customer side towards the back end to the very deep details into the databases and all the servers. And I remember some of the dashboards, you obviously were very interested in crashes, right? If it's a new app, you want to make sure that you're getting rid of crashes as fast as possible performance i assume is also a thing that you look at anything else yeah um crashes and and errors are really really important however we were able to basically monitor all the crashes but um but but those are basically not i mean not all of those are affected by the customers.

I mean, so therefore we were able to, or we are able to capture every minor problems that the developer guys later can correct on. So this is a really helpful feature and also our business guys are really like Updates index which is which is basically calculated by by by by Dynatrace Because because you know business people okay, okay, response time, errors, all this technical stuff, they don't really understand this.

But they have some indicative number, which is really helpful for them. And they can understand this. So this is one of the good measures that we can rely on. Yeah, that's cool. So AppTax, for people that are listening in and have never heard about AppTax is the application index where we basically look into user experience from an end-to-end user journey perspective and say that this user have and to make it very simple a good experience, a medium experience, a frustrated experience and we use a number in the end between zero and one to indicate how good that session was.

And I completely agree with you. It basically brings all of the important information together into one single number. And you can do a green, yellow or red. Yeah. And for business people, it's just really helpful because this is all technical stuff. So it's understandable that most of them, most of those information is not understandable. But there are certain parts which are really helpful for them as well. Now, in the initial rollout phase, when you started with this app, and I think I remember the dashboard,

you had many new versions that you then brought out because you wanted to then fix things that you saw in that process did you have daily meetings with your developers where you showed them the data that you that you've observed how did this yeah yeah basically during the initiation of the of the new application we have basically or we had basically a daily war room where every participant was on the table, basically behind the table. So business people, developer guys, product owners, operational guys, so everyone who's taking part of the project was there. And I think that was one of the key of the success.

So, yeah. And also we have defined alpha testers, beta testers. So within our colleagues. So we had the opportunity to test our application in-house first and then just release to the public. That's great. I hear this a lot with other organizations too,

like a friends and family thing. Yeah, that's great. Hey, there was another interesting thing in your presentation that was really interesting for me. There was a section around chaos testing. can you fill us in a little bit more because I remember hearing something about that this is even mandatory in the banking sector in Hungary yeah yeah it's we call it DRP

and it's it's simulating basically a disaster when some of our data center goes off. And then what happens? And this is the test that we are trying out our services, what happens in certain scenario. And of course, the main purpose is to maintain the business continually. So does this mean you're shutting off half of your infrastructure? Yes, definitely yes. At a lovely Saturday night we just kill it and then let's see what happens. And of course, everyone, I mean, every section for every services, the responsibility, the responsible person for every services

are ready at that night and they are ready to test it, whether is it working properly or not. And they are basically creating reports. And then afterwards, we are checking those reports. And then afterwards, we are checking those reports. And then if any kind of correction is needed, then we will just simply correct it and modify. So yeah, this is some kind of mandatory stuff. Yeah, nice to hear. Hey, I got one more question for you, because you mentioned in the beginning, the cloud was a big topic, AI was a big topic. I remember one presentation on AI and large language models. Is this a topic for you as well in the

bank? Definitely. Here in KNH we have Kate. She is the digital mobile banking assistant, personal one. And the solution is also located on your on your mobile app so it's hardly linked together it's also based on a language model the Hungarian language is basically developed by us locally but the whole program is driven by KBC globally. I just learned that yesterday at dinner I did not know, I knew that Hungary is not a big country but the Hungarian language doesn't really go beyond Hungary a little bit more but like there's not many people in the world that actually speak Hungarian and that's why and it's also very tough language yeah yeah yeah basically this is I think this is the one of the most difficult part of the

language language part here in Hungary because we don't have that amount of training data for our models because of the because of the of the language let's call it limitation because because it's so vocal. Yeah, it's interesting because I never, I mean I come from a German speaking country, so we have enough German material and obviously English is clear, I never thought about that there are regions in the world with languages where just the amount, the lack of the amount of material could be a problem to really train twin systems yeah yeah yeah but i i do believe that kate kate is really is really ready to to serve

our customers in a very proper way and also she is uh she's developing uh over the time so she knows more of our services already that we are offering so for instance you can make a transfer by just guiding her and just telling her that okay let's make a money transfer to my mom

and she makes it Can I ask her to transfer money to me as well? That's cool. Good. Hey, Adam, thank you so much for sitting down with me. You're welcome. And thanks for hosting us here in your office, for talking a little bit about observability.

And, yeah, I hope to see you soon and all the best for the future, for your future upcoming mobile releases for for Kate and everything else that happens okay thank you so much thank you