Georgie (00:00)

We start with AI Hack of the Week. Okay. Would you like to tell us a hack that you have?

Rada Stanic (00:06)

Anyone who knows me knows that I love doing my certifications, technical certs. I'm absolutely addicted. I find it as a way to, you know, force myself to stay current and learn new things. So I had to renew three of those and it kind of major work normally. So what I did is I used AI as my study buddy, you know, anything from helping me quickly grasp complex things and concepts to generate questions for me.

to practice, to give me concise explanation on things. And I managed to really pass the three in a record time. I renewed them like within space of two weeks. There's normally it would have taken me all over a month. So there is that and then I have another one. I know many ⁓ companies, many applications have these.

⁓ you know, applications that help you, you know, create professional documents and so on. So I have to do in the early part of the year, lot of planning, creation of strategy documents, what am I going to do with this and that? So we have this service in AWS called Quick Suite that allows me to organize all of my documents, something that's confidential, but also publicly available information. I provided a template of what I want included.

and it creates really solid first draft of something. So I've been doing that. And yeah, so that's what I would classify as a useful hack. And then there is the experimentation where I go and try, but maybe not. Yeah.

Georgie (01:46)

got a few of those as well where it's like fun to pick up, but then it's like, I'm not going to be doing this because I'm my daily. I'd love to ask you a little bit more about the learning and why do you think you are faster with it with AI? it because it's catering to your learning style? I'm a very visual learner. So someone explained a concept to me, especially a technical one.

find it a lot harder to ingest that, but if I can see it, even better if it's a YouTube video, something like that, I find I pick it up faster. What about you?

Rada Stanic (02:19)

So yeah, I'm the same, like the visuals are big thing. ⁓ And there's also hands-on elements trying to build something yourself. It depends very you at that sort of learning stage. In my case, I, for example, had these credentials already, so I needed to kind of conquer the delta of what's new, what's happened over the last year to build up that knowledge. And you can go the old way of reading documentation, ⁓

you're trying things by watching long videos and so on. But if you really have a solid base, and then you go and you ask, can you just explain to me how this or that works? And then you get the explanation, you say, well, you really didn't explain this very well. What do you mean by X? And very quickly, it really distills things down for you.

in a way that's easy for you to understand and I challenge it and I say, no, I don't think you're right here. And then you kind of go into that whole sort of argument and debate until you get an explanation that really resonates with you. At times I tend to forget that it's not a human on the other side. I love my debates.

Georgie (03:33)

Yeah, I kind of have that back and forth as well. I'm good friends with my AI though. I feel like we're like business partners almost.

Rada Stanic (03:41)

Yeah,

you know, it is, you know, and then you create a little persona that works for you as well. It's like, you know, sometimes it gives me the options and I say, ⁓ don't give me so many options. I want your firm opinion, the one that's the best. And it's good fun. So I think, you know, I call it my hack because anyone from primary school kids to high school to uni to a professional can...

use it in a safe way to just make themselves quicker and more productive.

Georgie (04:16)

Speaking of uni, remember when I was, we both did engineering. I remember the verbal. We had a lot of international professors, which was brilliant. But sometimes the accent would mean that my notes were wrong. I would write down the wrong thing. Like I'd be in calculus class and I would write the wrong thing for maybe half the term until I realised looking at the textbook, my goodness, I think they meant that. And then

all my notes, I'd be like, my gosh, all my notes are wrong. I really wish I had AI back then just to kind of sanity check that kind of thing. Take us back to where you studied university. us where you did it and why you chose engineering. I'm often asked why I chose it. I'd love to hear why you did.

Rada Stanic (05:00)

Yeah, so it's interesting. I studied in ⁓ a city called Sarajevo. It's famous for two things. It hosted Winter Olympic Games in 1984. And then it's famous for a very sad thing, the civil war that broke there in 90s. So I studied computer science and telecommunications. I was really ⁓ torn between wanting to be a writer or maybe be a mathematician, really.

But back then in a country that really admired and encouraged everyone to become a doctor, lawyer, or an engineer, it's kind of my choice was influenced by what is going to be the practical thing that I still love and enjoy. And I loved maths. I really, it just shaped my life very early on.

Georgie (05:37)

I wonder what that's like. ⁓

Rada Stanic (05:54)

So I thought let's study ⁓ engineering and that's how I ended up with computer science and telecommunications.

Georgie (06:01)

⁓ could be speaking deep into my soul. Mine was very similar, not writing or engineering, but music or engineering. ⁓ What ended up making you choose the math route over the writing route, do you think?

Rada Stanic (06:13)

I think it was a little bit like, oh well, you know, if you end up choosing writing and study literature or something like that back then, it was like, what are you going to do for a living? So you can be a teacher, professor, or really back yourself up that you're going to be a famous good writer. So I thought that studying something related to maths will give me more.

more choices, more freedom around what I really will end up doing down the track. And I mean, I don't regret it really. It's been a right choice in many ways.

Georgie (06:52)

Yeah, think about kindred spirits there. When I was studying it, was 20 years ago in Queensland and I would often get told, are you in the wrong building? I don't know if it was a little bit nicer to you in Bosnia or not. Like how did people treat you when you were walking into the classrooms?

Rada Stanic (07:11)

So it definitely was in minority, you know, and it was, I wouldn't say like that people were questioning or, you know, that it, but it was a minority of girls, women then. And it did shape one aspect of my life, ⁓ you know, through the entire life because it was just so few of us really. had one best friend and maybe a few other girls.

I've learned how to adapt to the way that guys are thinking. I became one of them. I then ended up building some really strong friendships with men, maybe rather than, you know, women because of that, of how that sort of happened. ⁓ So, yeah, I think that way it impacted my personal life and it maybe wasn't even at work until eight years ago at AWS when I saw a little bit more diversity, really, and then I had to adapt.

how do I work with them because I'm rock star?

Georgie (08:08)

I wouldn't know how to work with a female dominated interst after all these years. Same, same. ⁓ Yeah, agree. Yeah, I remember having a good posse of girlfriends, but made lots of guy friends. I went from an all girls school and I often tell people it was like going from an all girls school to an all boys school. But if there's any women listening that would consider it as a career, I would say not to let that deter them because I really enjoy doing it.

Rada Stanic (08:14)

It just saved me in that way.

Yeah, it's just great. You know, I think at the end of the day, you've got to choose what you love and you will find a way to fit in and get benefits from that. didn't feel that it kind of, you know, had any negative consequences whatsoever.

Georgie (08:54)

Yeah, agree.

A message for women wanting to delve into AI and technology further. Absolutely. Okay. I would love to talk about the time before you were at AWS and you worked at Nokia. Please tell me you were there when they had those indestructible 3310s and like tell us about what you were working on then.

Rada Stanic (09:13)

So yeah, was my, have very fond memories of that time. I came to Australia just having finished university and it was my first graduate job. It ⁓ was then called Alcatel acquired by Nokia. I started as a software developer ⁓ in what was effectively going to be telecommunications kind of related solutions. So really good combination of what I studied. So I wrote software, I tested it and then I went around the country

the team of people installing that software in Telstra exchanges. So the adrenaline rush at midnight, you know, making sure that that exchange wakes up with new software and you can make an emergency call in a minute or whatever we were given. It was a kind of full life cycle of you write it, you test it, and you go around the country and make a difference in people's lives effectively. ⁓

And that was my first job. And yes, those indestructible phones appeared at some point and it felt a bit magical. With some of these technological advancements, there are some that you go, yeah, okay, that's nice. And then there are ones that do change your life. And those phones replacing the big bricks and so on, they were good. They were handed to us to use and work with.

I don't always say I remember the time, say it's before my time, but it actually, yeah, I remember.

Georgie (10:43)

I even remember I used to ⁓ steal, admittedly it was my mom's phone and play snake on it. It just kind of blew my mind, those phones. They were so cool. Do you miss them? Would you take it at 3310 like a dumb phone now?

Rada Stanic (10:58)

What? It could have its seascape.

Georgie (11:00)

Yeah. Sometimes I wish I had a dumb phone. Like I need to be helped from myself. My son just started nippers, like the, the surfly saving. was, think the, like I had an hour and a half on the beach where I couldn't pick my phone up or look at my phone. And I remember afterwards being like, I think I, I think I need to do that more often, just like fully disconnect. Sounds like nothing an hour and a half, but I think.

Yeah, that's a long stint for me to not look at my phone.

Rada Stanic (11:30)

It

is. Yeah. It's like, don't know if you notice when you go on holidays or anything and you try and disconnect, you kind of, I become very kind of nervous first couple of days. I feel like something's missing. Where's my right hand, my phone. You just tempted. And then when you manage to disconnect, it just makes you so much more powerful, stronger, productive when you get back to it.

Georgie (11:53)

I agree. was almost like the first 30 minutes to 45 minutes were the hardest. And by the one and a half hours, I was like, I don't actually think I need it. Okay. So you've been in the tech industry for quite some time. When did you first look at AI and think this might be a bit of a game changer?

Rada Stanic (12:16)

Yeah, so ⁓ I actually think there was really a point in time because I remember I worked on some cool and exciting projects with machine learning, know, like highlight probably of my entire career, maybe would have been implementing fraud detection with Qantas loyalty for frequent flyers, you know, but it was quite, you know, heavy on the engineering side, being able to build, deploy, train models, have your data ready.

I had to get a good trusted outcome. And then I remember Chet GPT happened and I looked at it and okay, this is interesting. I think that was maybe around December 2022 or something, but then in 2023, the early part of that year, and I was thinking and kind of planning what my year might look like at work, what kind of technology I need to focus on for this.

clients I work with, which are predominantly a large enterprises. And I was really gearing for another year of like building this modern data platforms. Everyone was interested in, in, in getting more insights from data. And I thought it was going to be more of that. And yeah, there is this AI thing happening on the side, but very quickly it became clear to me that all that anyone wanted to do was just use then GenAI.

generative AI to build something. And there were people and companies who were quite smart about it that said, okay, I have this challenge or this opportunity and this technology generative AI lends itself nicely to solving that. And that was a good move, but there was a lot of hype as well around, let's just use generative AI to do something. And that kind of, you know, failed pretty quickly.

But I would say from 2023, there was no looking back. And then we were all, as you know, hit by there is LLMs and rich LLM and then there is agents and rich agent and concepts that just don't stop coming our way. So I think that was the turning point and it became very clear that that's the kind of one of those technologies like internet that's going to dominate our lives.

for quite some time.

Georgie (14:41)

I getting your perspective, Rada, because, you know, I've been in tech for 15 years and I do, and I say on the show often, like, I remember the waves of Web3 and, you know, cryptocurrency and the metaverse and these things that kind of flash in the pan and then disappear. Internet was a little bit before my time, but also not. Like social media kind of came out and there were these things that you just kind of can't turn back from. Yeah. And I love hearing your point about

going from data analytics being huge. I, I, in COVID I did a masters with the double major in data analytics. you know, I was hitching my ride on that boat. was like, data is going to be it. And then AI came in and it was like, just completely like tsunami everything else.

Rada Stanic (15:28)

That's the right word. That's the right, you know, I was like similar to you. Obviously I didn't do masters, but I was like data analytics was my thing. I was building my technical depth in that, you know, building lots of platforms with customers in real life. And then, you know, it very quickly became that's just a supporting act in this whole AI revolution. There is also a good thing about it. So, you know, like how I mentioned.

detection and that project, it was really heavy on maths and understanding the models and, you know, needing to be a good data scientist and good developer. But now I feel like technology is much more accessible to everyone and you can do a lot more with it without necessarily having to have PhD in ML and AI. It's really nice if you do, but...

Yeah, I feel like it became more accessible to more people. So that's one of my biggest positives, I guess.

Georgie (16:31)

I agree and I love that about it, but you have the computer science background. I technically don't. Do you think that there's something when you see people vibe coding and using tools like lovable and replet, is there something that you wish we would also do at the same time or kind of just play and figure it out?

Rada Stanic (16:51)

You know, it's, is place for everything. And I also do like, for example, I just explained how my roots were in software development and writing some really heavy code and fixing things very quickly and so on. I think there is time and place for vibe coding. And then what we also call is spec driven development, being more serious and intentional and, you know, focusing on production grade of creating some solutions to complex problems.

I think that both options have time and place and depends on, you know, your profession and what you're trying to do. So I do quite a bit of vibe coding myself just for fun, you know, creating some, you know, funny, quite mediocre games, but you know.

Georgie (17:40)

Hey! I'm about to see a...

Rada Stanic (17:42)

So,

you know, there is a lot to be gained from it. Just using things casual, like I have to create some quick demonstrations on the latest technology, wipe coding is great in that. But then if it's something more serious geared towards production outcomes, you want to really take that somewhere, then you go into using those coding agents and IDs to be more serious and intentional about the

quality of code that you're creating and how you can more easily extend and expand it.

Georgie (18:16)

so important for the listeners, Ed, see your incredible pedigree at AWS and your deep domain expertise and very sophisticated software background. And you're working with enterprise clients, enterprise ready, but to hear that you're vibe coding games and things like that, it's really empowering to people that might not have that background. So yeah, made, ⁓ listeners will know that I made a guinea pig of the day. Okay. Which was.

very important that your enterprise customers would be really interested in, I'm sure. ⁓ Tell us about something that you told me before you're really passionate about, which is AIOps. What is it? Why is it important?

Rada Stanic (18:58)

Okay, so I decided that maybe as this year goes on, there is a important and a serious discipline or an area to help enterprises solve. And many people will call it different names. I call it AI ops or operations, at AWS call it that. And essentially what it is like,

anyone can create one prototype, two, three, you might go and deploy some pilots in production and get some great results from that. But ultimately the question is, how do you scale your success? How do you create ⁓ replicable patterns that will allow you to build your AI agents or applications at scale in really fast, secure, predictable manner?

And I think that's a really sizable challenge to solve. ⁓ And that's why I'm passionate about AIOps and what it can solve. And essentially to just delve a little bit deeper into it. I always like talking about things as in there are three things to it or there are five things to it. for me, with AIOps, it comes down to sort of five key areas. Firstly,

Defining, like everyone talks about agents, like generative AI. So yesterday, so people are building agents, more agents to automate various things to solve problems. So defining the intent of your agent, the scope of what you really wanted to do, or as I say, creating and defining its personality. And then a second area, how do you solve for security, identity, authentication, authorization?

How do we separate what RADA, Georgie as a user can do versus what that agent can do? So that's a second sort of discipline. And thirdly, policy around those agents. What can they really do? And can you update that policy and restrict or give more freedom to that agent without having to redeploy it? How you can govern what they can do. And then another area.

which people don't maybe think about when they first create successful agentic systems. It's observability, monitoring, evaluation, you're experimenting, you're creating multiple versions. How do you evaluate which one was better and why? And lastly, in this day and age when we get new LLM every month or a week or a day,

Georgie (21:39)

Claude 4.6 coming out today.

Rada Stanic (21:42)

⁓ there you go. What a time. So how do you then manage that entire life cycle of which LLM you choose and how do you transition to a different LLM for whatever reason? So it's those kind of five areas that roughly describe what I mean by AIOps. And I feel it's really an important discipline and it's important to create that kind of framework and then become as an organization using AI. ⁓

Be more intentional, predictable around your success and what you can achieve.

Georgie (22:17)

Fascinating and you mentioned one of your enterprise clients in the past being Qantas, I'm sure that is the household names, right? Is there any area where time and time again you have to really kind of re-educate them? Like you talked about observability. Is there anything where you're like, it's our responsibility here to really highlight that this is important to them?

Rada Stanic (22:37)

There

is one area of servability definitely important, but there is an area where I feel like it doesn't even matter what the project or a problem statement or an opportunity to solve for is. There is a topic that comes up every single time and it's security. And by security, mean things that maybe we as consumers and people using AI in our everyday personal lives.

wouldn't necessarily think about to that level of depth. But really, I think, for example, imagine if you're a business, highly reputable business in Australia and you want to use AI, that you start asking questions, okay, where is this large language model actually hosted? Is it hosted in Australia? And then you say, well, yeah, we have models across all these reputable cloud providers. hosted in Australia. And the question becomes, okay, so what about my prompts? Do you store the prompts somewhere?

Are they encrypted in transit? What's the security measures around those prompts that are being sent to large language models? They shouldn't leave Australia. So there are those real intricacies around just because you can do something with technology, it doesn't matter that organizations here will do it until these deep and difficult questions are answered. And in fact, it...

really made me feel better working with a lot of financial services organizations. It made me feel good about the fact that I spent so much time on that because it means that my data, you know, what's related to me is going to be treated with data. It's like also, you know, financial services, they want to use these large language models to create all sorts of insights. But then just because I can doesn't mean that

Georgie (24:16)

Like your banking data? My banking?

Rada Stanic (24:30)

you know, personally identifiable information should be sent as prompts to models. So having to mask it, having to reduct some data before feeding it to a model and getting an insight or an outcome that you're hoping for. So I guess I talked a lot, but this whole angle around security and what happens to the whole data to model interaction and what is considered safe, it's quite a sizable piece of work for

Georgie (24:58)

Yeah, you're making me sleep a bit easier tonight too, thinking about banks asking the right questions.

Rada Stanic (25:04)

really

are and sometimes like, you know, you can relate to this being a techie effectively is you just want to really rush to create that cool outcome or application and you can, and I certainly feel, just why wouldn't we do this? It will work. The fact that kind of scrutiny is considered and taken before putting something in production, be a hundred percent comfortable and confident. ⁓

you're not gonna end up on the front page of AFR for data breach. It's a big thing.

Georgie (25:37)

Yeah, I unfortunately feel like it's a matter of time before something is on the front page of the AFR. Yeah. None of your clients, I'm sure. ⁓ Whose responsibility is it though? Is it the institutions themselves? Is it the tech companies? Is it both? Who asked the questions on security?

Rada Stanic (25:54)

It's really both, but ultimately those enterprises, those organizations creating these applications, they're custodians of their customers' data and information and they have to take on that ultimate responsibility. But it's also a responsibility of say, cloud providers who host these large language models and who host the data on behalf of these enterprises, but it's also thoroughly responsibility of those.

large language model creators of, you know, Google's and, and, and tropics and open AI and AWS and, all these companies, they also have to, so it's kind of like a shared responsibility model between all.

Georgie (26:38)

I could ask so many questions on this. One maybe ignorant question is why does it matter that our conversations are stored in Australian servers and platforms and things?

Rada Stanic (26:50)

Sometimes it's as simple as regulatory requirement. Sometimes it's just that. The regulator wants to make sure that data doesn't leave Australia. Sometimes regulator doesn't go that far, but that particular institution, organization has a business policy because that's how they operate. That's what they consider to be safe and secure. They decide that all of their deployments will be contained within Australia.

So it's a combination of factors. Not everyone's like that. There are other enterprises, especially like some of the SaaS application creators, they're quite comfortable operating globally across different regions, considering security measures, but they're fine with having global presence and operations. So it depends on the industry largely.

Georgie (27:41)

Interesting. had Tom from Heidi Health on the show earlier this year and yeah, very heavily regulated any healthcare data because it transcribes the doctor patient conversations. all Australian providers and servers and things like that. Where in the AI tech stack though, do enterprise customers like need things addressed time and time again? I'm curious if there's time and time again, you're like, this is where they

Rada Stanic (27:57)

Yeah.

Georgie (28:09)

need to be kind of more in the future.

Rada Stanic (28:12)

Yeah. So just interestingly talked about security. It's a big one. Another area for enterprises. And again, there is a difference in maturity and understanding between them and then between different teams within the same company is around power of data. A lot of people tend to think AI and LLMs, are magical. And they almost put these expectations that it just works. And they forget that

that what really will create those ultimately great outcomes is the quality of data that you make available to agents, to models and so on. If you don't have data that you can track properly and have its quality and lineage and control, fine-grained controls implemented, you will not be able to create great AI outcomes. So I find that

after security, which is just a foundational, making sure that there is quality and right data available is an area where I tend to spend a lot of time with those enterprises.

Georgie (29:24)

Guys, if you hear the words garbage in, garbage out, it's usually about data, right?

Rada Stanic (29:29)

It

really is, you know, and then also like, you know, remember, you know, the early days of AI and hallucination popped up. Models hallucinated. So the number one way to stop that is to make sure that the right tools, right information is available. Because if you have that, then the models can reason with quality information that they are provided.

Georgie (29:53)

Curiosity, is it always quality of data or is it sometimes there's not enough data? Is the volume ever an issue or is it more the quality?

Rada Stanic (30:00)

It's actually both, it's you've to have relevant data and it's got to be good, reliable data. So that's why, you know, we also ⁓ in, again, in the earlier days of AI, you know, I still remember like the reg and retrieval augmented generation, why it was so popular then it's because it was a way of enhancing the prompt with your personal private data as an enterprise, as an organization. And it automatically helped.

those responses to something that's relevant and accurate. Whereas if you just have publicly trained models, I mean, what can you get that's relevant to your business?

Georgie (30:40)

very broad-brushed.

Rada Stanic (30:41)

⁓ very broad brush strokes which are just not enough.

Georgie (30:45)

Yes, and we've seen scaling laws improve the volume and quality of inputs and data into the models. And then it seems like magic, but it's not magic.

Rada Stanic (30:50)

Yes,

It's not. lot, you know, and there are, depending on what you're trying to achieve and do, there are like techniques that you can use. Sometimes using an off the shelf model with, in combining it with, you know, your data being referenced through tools. Nobody really uses RAG as such anymore. It's more like your model has ⁓ access to the right tools. That's sort of one aspect.

But it's also sometimes you might have a need to fine tune the model, fine tune one of those large language models by using your own data to effectively create a new custom-made model. You might have that need. So those possibilities, depending on how niche the use case is and the problem and the challenge you're trying to solve, there are different pathways that you can take.

Georgie (31:48)

Do you ever do that with your customers, like custom models?

Rada Stanic (31:50)

Yeah, they're doing sometimes they might have their own models and they say, I have this model that really works for me. might've developed it on the side in my own premium environment or whatever. And I just need to import it into the cloud environment and then start making API calls to it. So all sorts of possibilities from fine tuning, importing your own custom models to sometimes.

You know, I call it practicalities of life ⁓ start to kick in when you might have like great outcomes using one of the sophisticated models and you say, this is great, but it's actually costing me a lot. And there is a bit of latency here because, you know, models taking its good time to reason and provide a response. So you might just distill that model down to something smaller. That's.

going to cater for the problem you're solving, but you're going to get better latency, better performance, better cost. So there are all these possibilities around how you can interact and work with the models.

Georgie (32:56)

What kind of business doesn't need that kind of level of compute and cost and model size? Like who would you kind of almost prod a little bit to think about having their own models? Like if there's an industry or a business size. ⁓

Rada Stanic (33:10)

I think it depends on the ambitions and the kind of like also how technical an organization is. Like in the early days, I remember what was really publicly known. Bloomberg decided to create their own model. also remember LG ⁓ did the same thing. So if there is a need, if there are people that can work on that, ⁓ it can be done. ⁓ But just because you can doesn't mean you should.

because it costs people money and so on, I would say 90 % of the time, the models that we have today with decent exposure to external tools, databases, systems you have, it can be solved by choosing the right size model for your use case.

Georgie (33:57)

You and I probably see this a lot, but I've noticed engineers can over engineer just because they can and it's exciting and it's fun, but then it's a little bit over-aush and it's like focus on the customers at the end of the day, right?

Rada Stanic (34:11)

That's right, know, and sometimes like the great models ⁓ turn up and people get really carried away by how good they are. ⁓ And customers go and try them and they say, it really solves my problem really well. And very quickly conversation moves from that. Okay, I've solved this problem, but can I now do it more cost effectively and with lower latency? And then I end up choosing, you know, a smaller model to solve it just because it's

makes practical sense.

Georgie (34:42)

Pick your brain on that for eight hours, Rada. That's incredible. ⁓ One quick question before we get to the predictions. What do you say to enterprises that are worried about letting their employees use AI?

Rada Stanic (34:55)

Yeah, I really say that it should not be even a discussion point. I mean, we're well past that turning point or should be, is it the right thing? I think the question becomes around how do I empower my people to be able to experiment safely? How do I create my responsible AI policies and encourage people to explore the tools and see what works for them?

and encourage them to upskill and find that medium of where they are more productive and they develop their skills and do so in a safe way. I think stopping, banning, prohibiting, we're so past that. So yeah, there's no place for that sort of conversation, I think, on not encouraging people to do it.

And there are so many areas where it's already well established, like with coding ⁓ and agentic coding companions from anyone. At AWS, we encourage to try different things, but there are at the same time policies and reminders around what data you're using. Do you know what's happening with that? There has to be an education around whole operational and people side of things.

as well.

Georgie (36:19)

Yeah, at Google, even copy pasting into certain models like error codes. And it's like, no, that's just from a WhatsApp message. I swear it's fine. But I get why the policies exist. And I completely agree with you. find it kind of horrifying to kind of keep people back from learning how AI works. That really frustrates me personally. Yeah.

Rada Stanic (36:42)

Definitely.

Georgie (36:44)

What do you think our next generation, you've got children, right? I've got children that are quite young, but what do you think they'll be greater with technology in this new future and what can maybe concerns you about the next generation?

Rada Stanic (36:58)

So I'll start with the concern first actually. The one thing, and I don't know, you might have the answers to this and other people might say, what are you talking about? But I am actually quite concerned because I can't see it clearly around what is going to happen with the junior roles. So my daughter is at university, just started. And whether you are in finance, banking,

whether you computer science, any junior role. ⁓ There used to be this period of learning, you're given simpler tasks, simpler jobs, and you do those and as you do, you progress and then you become this super senior expert that skies your limit. So there was very well oiled path from a graduate to developing into an expert in your field.

I feel that AI is rapidly filling that space. I mean, we've all seen the stories, JPMC, many large banks, they create agents who are junior banking analysts and so on. Coding tools are becoming more powerful. So I'm really not thinking that there will be no need for.

senior bankers or senior software developers. think it's going to be how do you get there? So that's that, kind of transition from junior to senior, I feel it's getting affected more increasingly by AI. And that worries me a little bit how, that's going to be playing out. But in terms of a positive for this sort of young generation, I think, you know, there is an opportunity to, for them more than ever.

to be able to really handle change much, much better than we ever did. Things are evolving so quickly, so rapidly, and they are trying and experimenting and doing things at an increasingly fast pace. So I think they will be really equipped with that ability to adapt and adjust and maybe become more entrepreneurial also and put their creative ideas in.

practice much better or quicker than we ever did.

Georgie (39:22)

When you think about this generation, what it's been through with like COVID and everything they've been through, if that happened, like, gosh, I think I've experienced my first change at like 28. Yeah. Yeah. Yeah.

Rada Stanic (39:37)

Exactly. if I reflect on my career as well, I use this analogy with my friends. know, I've been a technologist my entire life. I really love technology, especially now the whole intersection of, you know, technology and business, solving business problems with tech. But I love learning new things all the time. And there is hardly any technology I haven't worked with, to be honest. And I remember there was time you'd learn something.

And then for next three years, you're good. You just keep being awesome at that. And then that got shortened. And there was a period of time where I've never experienced not knowing something. And I remember AI brought me this experience. I clearly remember I was on a bus one morning. We were working on this AI ⁓ chatbot project with one of the banks. We working with Anthropic Claude. I was super comfortable. Things were getting good.

And this customer calls me, Meta has just released this llama model. Maybe we should use this. And I'm like, what is that? When did that happen? I've learned to accept that there are things that I don't know, that I'm not aware of. It's happened, like, you said yourself, all 4.6 is coming. So it's like I had to become used to this shortening cycle of how long

your knowledge is effectively valid. And this generation is just exposed to this completely different level. So that's going to be their strength.

Georgie (41:13)

You have crafted the perfect picture for how in my gut I'm always like, my gosh, I'm not keeping up. I'm not technical enough. I'm not entrepreneurial enough. And I used to think I was ahead of the curve. That three year gap, was like, good luck catching up suckers. And I'm like, God, everyone's ahead and not ahead.

Rada Stanic (41:35)

And

now that's a kind of like skill, personal skill that I've embraced and I say, okay, I will be okay with saying, I don't know, I haven't heard about that. And then I'll quickly research and find out and form an opinion. We live in such time.

Georgie (41:50)

Yeah, I did this big LinkedIn announcement this week about this is what I'm leaning into is this messy middle. There's the builders and then there's the people that kind of just are figuring out what AI is. And then this messy middle of like, how do we kind of navigate through being more proficient than just a chat bot, but also not necessarily being a coder? How do we work in that space?

Rada Stanic (42:15)

I think it's great space to occupy.

Georgie (42:18)

It's scary also. ⁓ Okay. You've touched upon rag. Is it yay or nay for 2026?

Rada Stanic (42:26)

I think, you know, it's more of a nay for just, it won't be mentioned as such. The concept is still inherently there, but it's more about like, well, LLMs will just retrieve this information and, you know, it won't be like referenced as a fancy retrieval, augmented generation and use in a way where we feed information to a prompt. Models will just access the information via tools and get what they need to provide the output. So yeah, I would say.

Georgie (42:54)

I'm glad you said that because it was on my out list for 2026. And what about MCP? MCP was something everyone wouldn't stop talking about, I feel, four months ago.

Rada Stanic (43:05)

Yeah, it's kind of because the promise of MCP is quite a big one, you know, especially with enterprises, they don't like any bespoke integrations and custom interfaces. And I mean, who does? So MCP promises to solve for some of that by providing that sort of unified access to external tools, databases, APIs. So again, I think it probably won't be mentioned as much, but will be used in the background as a given. It's just like.

you know, and you use to create cloud applications. No one would ever talk about how you designed for networking around that or how you designed for security around that. It's just something you had to do and you focus on what the application does. So I think MCP will become just that building block around how you in a unified way allow large language models to access the tools.

Georgie (43:58)

Building block, not marketing.

Rada Stanic (44:00)

Not marketing, like it used to be like people stopped talking for a period of time about LLM, so it's MCP, MCP. ⁓

Georgie (44:07)

Yes, I remember that. We are at our rapid fire. If you can, 15 second answers for the last questions. Are you ready to go, Rada? Can you confirm or deny there might be an AWS data center being built in Australia?

Rada Stanic (44:17)

Here we go.

I can't confirm or deny, but I can tell you there is $20 billion going in the infrastructure investments by 2029. So it will be chipset services for AI and all of the cloud infrastructure. I'll leave it.

Georgie (44:41)

Not a small amount of cash. What do people get wrong around the competition between Microsoft, Google, AWS? We're friends, right? And we've worked at different tech companies. What's your take?

Rada Stanic (44:55)

People tend to think it's this fierce competition where we look over the shoulder of each other and what each company is doing. And I think in reality, they're all great companies doing great innovations. And certainly at AWS, the focus is on what do customers want? What do we build for them? And what can we innovate with that our customers do not even think about, but we know they will need. So I think it's, if the end result is great, everyone innovates, creates

fantastic outcomes they can use to do great things for business. it's focused more on the customer and the businesses and what they can do rather than on each other. think we focus on each other too much.

Georgie (45:38)

Yeah, it's not personal, right? Yeah. And I love almost like the competition means that just each company keeps building the most incredible product. Yeah, I love that. Are you more positive or negative about the idea of having a robot in your house, Radha?

Rada Stanic (45:47)

Absolutely.

So I am more positive, but I do have strict requirements of when I would let them in my house do specific things. Lawn mowers, vacuum cleaners, okay. Venturing beyond that, I would love them to prove their value to me before I let them into my house.

Georgie (46:15)

Great answer, I feel the same. Not near the chickens though, right? Yeah, keep them away from the chickens. Finally, will AI give us a four-day work week anytime soon?

Rada Stanic (46:27)

Definitely don't think so. And I'd love to be wrong. I'd love to be wrong, but I don't think so. I think all it will do is create higher, greater expectations and all and any productivity gains that we get, which we are already and will in the future, it will just raise that bar for more expectations around what we can do and achieve.

Georgie (46:52)

Sadly, I'm not about to start blanking out my Fridays for gardening. Yeah, yeah. Thank you. Rada, this has been such a joy. How can people follow you? it LinkedIn or would you like to say to the listeners?

Rada Stanic (47:06)

So there is LinkedIn, know, I do keep in touch with people there and, but also probably the best place to find me is I love my public speaking. So anytime I do an exciting project or, you know, I learn about technology, I will be on the stage somewhere telling the world about it. So you can meet me in a lot of industry events explaining how technology helps business.

Georgie (47:33)

any big 2026 events we can see you at that we know about or.

Rada Stanic (47:37)

So there will definitely be a Sydney Summit for AWS every year. I talk about something that I'm passionate about usually data or data and AI. Last year I did a big ⁓ keynote on our Innovation Day. So yeah, I think that's where I'll likely meet.

Georgie (47:56)

⁓

A people, you know, just casual,

Rada Stanic (47:59)

It was one of my personal highlights, you know, speaking to such a large audience and providing some value, hopefully. So that was an exciting event for me.

Georgie (48:09)

Well, this has been a pleasure. Anyone that gets to experience you in the person, I definitely recommend. Thank you so much for being on the show.

Rada Stanic (48:16)

Thank you so much. The pleasure is mine as well. And I can't wait to see where your podcast goes this year. You occupy such an interesting space and I look forward to learning from you and your guests.

Georgie (48:29)

We've heard it from the expert guys. Thanks, Radha. Thank you.