Speaker 1: 0:07

Hello everyone, welcome to another episode of Microsoft Community Insights Podcast, where we share insights from community experts to update Microsoft. I'm Nicholas and I'll be your host today. In this episode, we'll dive into AI work in a theme of what does every developer need to know? Today, we have a special guest called Brian Swagger. Could you please introduce yourself please?

Speaker 2: 0:31

Yeah, brian Swagger, I'm a cloud and AI solution engineer at Microsoft. I've been with Microsoft for almost 10 years now. I live in the Tampa Florida area, so those in and around Florida, hi. And yeah, I've been a developer most of my life, so across a lot of different technologies, so all kinds of different technologies and love, you know, training and upskilling of individuals. So been in the certification space for about 15 plus years, as you can see, and I've also been a Microsoft certified trainer for many years, so I like certifications, but they just help to drive learning, so it's a lot of fun there.

Speaker 1: 1:19

Yeah, because I saw in there like you're like one of the exam development, so what does that mean?

Speaker 2: 1:28

Yeah, so I'm a content development manager. I've worked across a lot of different exams just either being a subject matter expert, writing items, as well as helping in the development of the exams. So I've been helping for many, many years just trying to craft the exam.

Speaker 2: 1:49

Long story short, I did an exam way back when and I had a challenge actually by a co-worker. They pretty much said that I didn't know my stuff because I wasn't certified and I just really felt that certification, you know at the time, wasn't as advanced or worthwhile. So I went and I got the certification it was a BizTalk certification back in the day and I actually aced it and I came back and said it was a waste of time because it was all. The questions were multiple choice, they were all A, b, C, d, you kind of guess. You know, eliminate two and guess between the other two, and I just hated it. So I put my name in to be a subject matter expert and in this database at the time it was a website and Microsoft had an app and I got chosen to start writing exams and I've been writing and developing exams ever since.

Speaker 1: 2:50

Okay, so you're one of the first ones to have created an exam and then improving it, like whether it's through the previous MCSA, mcse and Windows Server.

Speaker 2: 3:02

Correct. Yeah, yeah, the old MCSD and MCSE days. Windows Server Correct. Yeah, the old MCSE and MCSE days yes, that was the days that I was taking exams and things like that. And of course, now we have various numbers like AI 102 is the AI-focused exam, az-204 is the Azure developer, az-400 for DevOps, things like that.

Speaker 1: 3:26

So yeah, so at what point when you were developing exam, did you have a night? Did you guys in your team have an idea of making lab-based scenario? Because I'm not.

Speaker 2: 3:36

Currently is no lab-based and case studies yeah, yeah, I've always wanted case studies and labs. So my thought process there you know worldwide learning decided to make it more hands on and engaging. I've always loved that concept. So I'm a certified Novell administrator from back in the Novell days and even then, when you got a Novell certification or even the Cisco exams, you actually have to get hands-on with the technology and that's what I loved about this. The whole lab experience is you actually have to do something. It's not just a question and regurgitating knowledge, it's very actionable and actually having to do something. So, and the fact that we now have Azure to be able to host those labs you know we have Azure Virtual Desktop, avd and Azure Infrastructure to be able to host you know those lab experiences. So that's, I think, yeah.

Speaker 1: 4:37

Yeah, because it's like fast forward to the present days. Now we have Applied Skills, which is quite a step up towards the exam, which is more of a lab-based scenario as well. So once you do that, you get certified. So you know what to do, you have a production environment, you know what to do in a real-world scenario. That's quite cool.

Speaker 2: 5:02

Exactly, yeah, and those are all done through Microsoft Learn. So if you go to microsoftcom slash, learn all kinds of free labs and modules and stuff. That's the other thing. The team expanded and has offered a lot of free learning in that area, so that's what I really like, yeah.

Speaker 1: 5:19

Okay, so let's dive a little bit into your current role as a principal solution architect engineer. So what is your day-to-day comprised of?

Speaker 2: 5:35

I am in a line to the retail and consumer goods area of Microsoft, so I work with a lot of retail and consumer goods organizations so a lot of quick serve restaurants, a lot of retail and consumer goods uh organizations so a lot of quick serve restaurants, a lot of grocers, big box, so any any of those areas, uh, that you've heard of you know global brands, uh, global, uh, like I said, quick serve restaurants. So I work and I help solution, um, with our teams, uh, you know, in terms of helping them to realize the benefits of Azure, as well as, of course, ai, you know, and the benefits of using Azure AI Foundry. That's kind of my key focus is, you know, on developing AI agents that really affect things like supply chain, inventory management, some of those other key aspects that you know retail and consumer goods organizations you know are concerned with.

Speaker 1: 6:34

So, yeah, okay, yeah so you help them adopt into like AI and then using the latest technology, like AFR and G2C agents or MCP as well, yeah, exactly.

Speaker 2: 6:52

Yeah, essentially building proof of concepts, minimally viable product, demonstrating what can be done, the art of the possible, and showing them what they can do with the technology. To, to positive.

Speaker 1: 7:06

Is there any? Is there any? Do you see? What kind of like case studies or proof of concept do you see more from customers? Is it just like a child bug agent or that's to make the ai into the environment as a consumer?

Speaker 2: 7:25

so a lot of the consumer experiences now are becoming down to enhancing chatbots. I would say as kind of our bread and butter. That's pretty easy to do, chatbots. I literally can give you a GitHub repo that can show you how to pull down code that you can build your own chatbot Chatting over data, providing it data context effectively in terms of doing augmented generation of the output and looking at that. Those kinds of things are the default. I would say where we're seeing advancements is building multi-agent systems that affect supply chain or inventory or key business aspects, even content generation for marketing departments, generate images and copywriting and things like that.

Speaker 1: 8:22

That's where we're seeing the key benefits is reducing individuals time, um, to help out with those kind of things yeah, because I guess one of those, one of the scenario could be like helping them renew their contract for suppliers and stuff every year instead of manually after school. Then you can get agent to do it for you yeah, exactly yeah, and you can.

Speaker 2: 8:43

You can have any. You you know developing agents, you know that communicate with one another, or agents that act on behalf of you know individuals or processes. So what's nice is you know we have security around agents that can identify themselves. So they're you know, we understand, from an auditing perspective, exactly who did what. The agents are actually a who from that security perspective as well.

Speaker 1: 9:13

Okay, so there's different type of agents and you audit the agent as well as monitor it to see what type of information is put into the agents.

Speaker 2: 9:25

Yeah, exactly. Yeah, the agent can act autonomously, but still you want to make sure that it has its own security and it has its own guidelines and that you can audit and control the agent in terms of security boundaries. So that was one of the biggest problems that we saw with model context protocol originally.

Speaker 2: 9:45

Yeah, so you know mcp, uh, you know, was uh the you know protocol that was uh came out from anthropic and has been adopted by the industry. Um, you know, really to to give all these agents access to additional tooling. So we want to be able to plug in to other areas. Right, we want to access other you know systems or do things autonomously, but the problem was being able to authenticate and communicate securely with those systems. So MCP was actually ratified, I think, in April it's evolving so quickly, um, but it's. It was ratified in april to allow third-party idp so you can do um microsoft uh entry id or octa or other third-party authentication to be able to do um you know authentication authorization on behalf of an agent. So security was a big thing and mcp has came a long way with that too.

Speaker 1: 10:47

Plus, you don't need to be using a Microsoft product to use agent, because you can still use agent locally, like you have to find your local or build your own agent as well, so you don't have to be solely using Microsoft.

Speaker 2: 11:02

Yeah, I mean, obviously I'm a little biased working for Microsoft, I want you to use Microsoft's technologies. But yeah, I mean writing agents. You know, in terms of technologies, yeah, you don't have to use. What's interesting is most of the frameworks that are out there. Even Microsoft's own frameworks are open source. So if you're looking at, you know, orchestration, youration Symantec Kernel is an orchestrator for agents and it's open source. So it was birthed at Microsoft and it's continuing to be updated by the community, but it is open source software. So you don't have to leverage Microsoft-specific technologies but the ecosystem's there, including models. So models from OpenAI, models from Anthropic models, from XAI with Grop. So we have a multitude of models that are out there. Even Meta's Lama is out there in the foundry that you can leverage from Azure AI's foundry. We host just a plethora of models that are out there. So to me, models are interchangeable, those are constantly evolving and use the right model for the right context.

Speaker 1: 12:19

So how would you keep up to date with knowing what module you need to do, need to do what thing, and update the code, because every time it comes on change what's really nice is, you know, back in the day we're going to build, we're going to architect our systems with abstraction and interfacing right.

Speaker 2: 12:41

So we're going to abstract away essentially that model and being able to build a system that allows us to swap models out at any time, uh, so we want to build in those abstractions and connectivity accordingly. So I would would say, you know, when you're architecting your systems, be able to swap in and out models and make that, you know, plug in place. So there's always going to be newer models that come out that have new features. And you know we've got new features even now with model routers. So there's both in model routers and things of that nature that you can do abstraction. So's both in model routers and things of that nature that you can do abstraction. So you know, I would convince folks to architect with like APIs and API management and abstract out those model endpoints so that you can swap out you know your models and get a better model in there if it makes sense for the you know the business concern.

Speaker 1: 13:34

So yeah, so I guess one of the important things is that you have to be up to date to see when the new models come and consult, changing to meet your demand and be able to use it according to best practice and create use cases on the module itself. Exactly what are some of the best use cases that you see modules used so far?

Speaker 2: 14:02

Yeah, I mean, obviously you know you talked about like chatbots, so chatting over data, you know providing it, you know user manuals or what have you. But I'm seeing a lot of organizations build chat interfaces that you know are leveraged across the entire organization. I see a lot of those even surfaced up through Microsoft Teams. So Teams being kind of that area that everybody's already chatting with one another, creating meetings and things like that, things like that. So they're augmenting it with agents inside of teams and they're building their own that interfaces to their own company's data and kind of surfacing that up through teams.

Speaker 2: 14:44

So I would say that's the biggest thing that I've seen and that's just the chatbot experience. Imagine taking that to the next level and using these probabilistic model calls to do things like evaluating whether or not you know we need to affect supply chain in a certain area based on trends or what have you. So being able to do maybe some pricing variations as well based on trending data, based on trending data or some of those type analyses, has been very, very beneficial as well. So the key aspect has been using it in different business use cases where it makes sense. The biggest impact that I've seen, again, by default, the chatting experience, but I would say marketing and generating marketing collateral or developing new design ideas is also another area that I see a lot of use cases as well.

Speaker 1: 15:45

Yeah, because I normally think about using it when you can add values to a specific use case. Use case so, whether it's reducing cost, improving security, helping with code like get rid of your news code and stuff, repos and stuff Because I'm currently in my workplace and we're trying to think about using it, so trying to find ideas where we can use it, whether using it in Azure DevOps, whether using it in Azure for security purposes to monitor defenders of cloud and stuff. So always think of things that add values for your use case.

Speaker 2: 16:28

One use case. That's really neat that I see a lot of retail organizations developing and using a lot of these models are multimodal now.

Speaker 2: 16:37

So obviously they can recognize. You know, audio and video and pictures and all of that. What I'm seeing is a use case where you walk in to a retail organization that has products on shelves, right, you take a picture of that product, of that shelf, that shelf, and the model and the analysis can understand how many products are on that shelf. Um, you know, do inventory of that shelf and and make sure it understands, uh, the actual layout of the product and see if there's any competitor products on the shelves as well. So a lot of those retail use cases are already being developed and used in stores and by individuals that are stocking shelves and stuff like that, you know, to save time.

Speaker 1: 17:26

Yeah, so in your experience, what are some of the AI practices or things that any developer should know, like GitHub, codepilotilot, ai foundry?

Speaker 2: 17:38

I would say I mean, I'm a developer in many, many years. So, yeah, uh, you know I love ai foundry it's. You know the central area in terms of agent service. Uh, you know all the ai models and things like that. But you know we've got copilot. You know copilot's an overloaded term.

Speaker 2: 17:55

I would say so think of Copilot as the UI for AI. Copilot is there across a lot of different products. You've got Copilot in Excel and Word, but we also have M365 Copilot. That's effectively going to be able to do graph calls so it'll be able to read you know your email. You want it to reach out and read your team's chats and compile all that information together. So M365 Copilot's good for that use.

Speaker 2: 18:27

Then you get into Copilot. Studio is built on the low code, no code type platform, so it's built on Power Platform and you can leverage that to build agents as well. So you can plug into a lot of different systems. And then you get into AI Foundry where you're deploying the models, doing analysis, governance, evaluation, monitoring, tracing, all the fun stuff. I'd say, as a developer, we get our hands dirty, but that's where it gets really fun. But Copilot, as an overloaded term, I just want to make sure it's like the UI of AI, but it's iStick and Foundry. So for me it's leveraging custom code in semantic kernel, maybe as an orchestrator there's other orchestrators that are out there but leveraging custom code and I'm a C-sharp developer. I don't know Java or what language you know that you use C-sharp's my main one but you know developing custom agents with models deployed inside of Foundry.

Speaker 1: 19:33

Yeah, because I know that everything tends to now be in VS Code. So there is AR Foundry Toolkit, ar Foundry, there's AR Toolkit, there's probably C Sharp extension. So all you need to do is have VS Code really and all your and you just do what you do best like programs, stay in that interface mode and just maybe use Copilot to help you in agent mode or sometimes use AF Foundry.

Speaker 2: 20:05

Yeah, exactly, and you're mentioning, of course, from a developer perspective, the best Copilot, in my opinion, is GitHub Copilot. Yeah, that allows you to essentially do vibe coding, so prompting it to allow you to write unit tests or write code for you. So GitHub Copilot built into VS Code. But, to your point, we've got a lot of extensions in VS Code. Even Symantec Kernel has an extension out there. Github Copilot, of course, that I'm talking about the Azure AI Foundry extension and other extensions to help you build AI solutions and AI agents.

Speaker 1: 20:47

Yeah, so I know recently there is GitHub Beastmode for the latest VS Code Studio. Have you used it before? Have you used Agent Mode or the new Agent Mode, beast Mode, yep, and what's your opinion on it?

Speaker 2: 21:11

I think Agent Mode's working really well. I love the fact that you can choose different models in there. You know the I've been using the quad sonic models mostly. I see a lot of success with the quad sonic four model in my experimentation. That, you know, it depends on the scenario. You know the best model for that. But yes, agent mode. Huge fan of agent mode. I use agent mode all the time, whether it be to, you know, like I said, write my unit tests, build MCP integrations. What's really nice too is it has MCP connectivity so you can actually use agent mode to write code that would say interface and do things inside of Azure itself. So we have an Azure MCP server that you can plug into and other MCP servers to do that connectivity to other systems. It works really well.

Speaker 1: 22:10

I think you wanted to show us some cool demos, brian. We got in anything really. So AF Foundry or the new GitHub Copilot, we're looking forward to it.

Speaker 2: 22:21

Yeah, absolutely. I can go ahead and show a couple different things, starting with, essentially, ai Foundry. I talk about AI Foundry and all the fun models and things like that. So let me go ahead and show my screen real quick.

Speaker 1: 22:41

You could tell us a little bit about Vive Coding, because I heard lots of terms on it. Everyone thinks Microsoft built this. I'll talk about Vive Coding, but from my understanding it's just a way of creating a product with using AI yeah, not other teams as quick as possible yeah, and it's you know, essentially using uh vs code.

Speaker 2: 23:09

Uh, you know, and doing, you know prompting, uh, prompting. So I mean, if you're looking at, you know here's VS Code. I have VS Code up and running. You know, down here in the lower right you know I've selected the Cloud Sonnet 4 model. I've got I'm using GitHub Copilot. This is effectively. You mentioned agent mode, so I'm actually going to change it for agent mode. I'm currently at defaults to ask.

Speaker 1: 23:37

You may have to zoom in a little bit, Brian.

Speaker 2: 23:39

Zoom in a little bit.

Speaker 1: 23:50

While he's zooming in, I've got a question how would someone know the difference between using different modules and the pricing stuff? Because someone can be using ChatBTP 4, but is concerned about it might cost more if using ChatBTP 5, the latest one.

Speaker 2: 24:10

Yeah, different models are going to cost different amounts, and it's usually based on tokens are going to cost different amounts and it's usually based on tokens.

Speaker 2: 24:17

So think of a token as the number of characters that you're going to send to and from the model. It's typically around four letters, so think of four letters as one token. And tokens are input and output from the model. So, as you're doing the communication with a large language model, you have input and output from that model and that is essentially how they're charged. But usually there's a large amount you might get. I don't know the exact pricing on all of the areas that I could look up pricing, but you usually get a lot of tokens initially, like a million tokens or something to be able to communicate with that model. But, yes, yeah, depending on how much you ask it and the the context that you provide that you're going to send it uh, you know size and the prompts, etc.

Speaker 1: 25:06

that's how it's going to determine how many tokens it's going to use yep, and you may need to upgrade from the free tier if you need. So, you may have to use the enterprise tier to get more of a token.

Speaker 2: 25:17

Exactly, exactly yeah, and you know, um, you even have access to gpt5, uh, which just debuted, uh, you know, recently from open ai. Yeah, I'm still experimenting with gpt5 to see what my, um, you know how my, how my prompts are working and see how good the responses are. But I mean effectively, you put it into agent mode and you're going to prompt it to do something so you could say, hey, effectively, build me a brand new application, aspnet, or create an MCP server. You could literally just type all those various things. That's what I was doing previously. I actually prompted it.

Speaker 2: 26:05

Help me build a Power BI dashboard that connects to Azure SQL and can you tell me how to set up an Azure function that processes data from blob storage? So you might ask it. You know essentially those things and it's going to literally think and write the code for you and be able to actually generate that code and add that code to your project, so it's able to code for you and be able to add that code directly right into your project as you go. So that's kind of how it works. And again, you can change models accordingly and see the different output that you get and the type of code that you get as well.

Speaker 1: 26:45

Yeah, so the information you get is based on how good your prompt is. So the more of your contacts you put in the the agent, the more of a result you get. And then as a developer, you would just be wary of your important information, like your secrets and stuff. So all of your code is like public code and it's a structure reading, so it's a context that you give it.

Speaker 2: 27:13

Yep, exactly, yeah, and you can add additional context. So if you had additional context that you wanted to provide, and again, you know you can even connect, as I mentioned, to various MCP servers. I've added in. I have an MCP JSON file, so I configured some of the MCP connectivity. So you know there's various MCP servers that I have here, including, you know, the documentation. So we have actually an MCP server that connects to. You know Microsoft Docs Playwright, if you're doing, you know UX testing, you know Brave Search. So there's a lot of different MCP servers that you can attach, connect to Again even Azure's MCP server attached. Connect to Again even Azure's MCP server. One thing I do want to note is we actually do have the MCP center Folks are not familiar with this that's mcpazurecom.

Speaker 1: 28:11

We actually I think that's new, because I'm not aware of that. There's a breakout repo. I don't know that there's a site on it.

Speaker 2: 28:19

Yeah, so this might be brand new to a lot of people. This is the MCP Center. This gives you. It's essentially powered. You can see in the upper left by API Center. So API.

Speaker 2: 28:31

Center is a new product, or newer product, I should say, is a part of API management. Api Center allows you to organize and catalog all of your APIs, but this is an MCP server. It shows you all the various MCP servers that have been published and you can access them. So you'll see a playwright, like I talked about for testing purposes. Here's Microsoft Learn, if you're looking to connect to documentation. Again, here's Stripe. You'll get GitHub, Azure DevOps, Azure's MCP server. Here's Box. So there's various MCP servers that are already here and published and, of course, you can develop your own MCP server, run it locally, host it where you'd like. But these are some MCP servers that are published out here If you want to connect and connect your agents up to them, including Atlassian here at the bottom. I didn't mean to.

Speaker 1: 29:29

Some of these MCP servers Microsoft MCP, like Atelier, boxer Swipe, so I think these are like partners MCP. But would you like to put their product here as well?

Speaker 2: 29:40

Yeah, exactly. Yeah, so these are partner MCP servers. Yeah, exactly.

Speaker 1: 29:47

It reminds me of like Microsoft Marketplace. Yeah, similar to that.

Speaker 2: 29:52

Exactly, except for MCP. Yeah, exactly so. Model context protocol and connectivity, from you know leveraging it, from you know Anthropic came out with model context protocol.

Speaker 1: 30:05

Yeah, nice. So I guess the best way is just to try on each of them, and then it will go to the GitHub repo and just install it and just play around with it and then work out which is best for your use case to use.

Speaker 2: 30:21

Exactly yeah.

Speaker 1: 30:23

Depending on your use case, because I know that there's quite a lot. There's even one for Docker as well. So, docker MCP yeah, it's quite a lot, yeah, okay. So that's quite a lot. Okay, let's talk about how you can secure an MCP, because I saw some information that you can actually secure it using APM management. Yes, that's correct. So how does that work using APM?

Speaker 2: 30:56

So APM management actually allows you to you know, catalog and connect to your MCP servers. Like I was talking about earlier, you can effectively authenticate, you know, with your MCP server, so when you're hosting it. We talked about third-party security or third-party IDP identity providers such as Microsoft Entra ID, so that it allows you with API management, so you're cataloging and referencing your MCP server, but you're also able to authenticate with that third-party IDP. So your identity provider is allowing you to do an intra-authentication before you actually connect to the MCP server. So your entire connectivity back to these MCP servers are fully secured. So the communication.

Speaker 1: 31:52

Okay, yeah, it's like an identity Okay.

Speaker 2: 31:55

Correct, correct. Yeah, we actually have an example that shows APM management with MCP integrations, so I can bring that one up too. I didn't have that one ready 100%, but I can bring that up too, yeah, but yeah, we have examples that show that integration and there's tons of yeah.

Speaker 1: 32:26

Today, when I went on API Management, I can see there's a feature called MCP. There it's like the built-in feature. So I'm not sure, because that's the way Microsoft is going to it. It's the code MCP. Now.

Speaker 2: 32:40

Yeah Well, yeah, it's API management has that so to be able to, you know, essentially expose and govern, you know, those MCP servers. So that's, that's the objective there.

Speaker 1: 32:53

Okay, that's brilliant. Yeah, that's brilliant, that's curious, okay, so let's talk a little bit about upscaling and stuff. So what's the best way to get started with all these agents and MCPs? Because sometimes it's scary because of the fast moving of AI every time you're changing new modules, so how would you keep up to date with those?

Speaker 2: 33:20

I mean, the biggest thing I can always tell folks is get hands-on. So, you know, get a free Azure subscription. Get hands-on with AI Foundry. You know, try out some of the samples as well. You know, github has a lot of example projects that are out there that can show you how to do a lot of these things, especially even MCP A2A, which is, you know, google it's agent to agent protocol, which the industry is adopting as well. So A2A you know there's a lot of great examples that are out there in terms of the projects. But another area is Microsoft Markets. So Microsoft I keep referencing that.

Speaker 2: 34:00

Microsoftcom learn. There's a lot of free resources out there in terms of learning modules to teach you all these various concepts. So no money involved. You just access that site, go through the various learn content. That's what I love about it. You can learn about all these various concepts right out there on the Learn site. Github, like I mentioned, there are various GitHub projects that are out there, even the new A2A NET SDK. So we have a software development kit for NET developers around an agent-to-agent protocol. So it's a NET library that helps you to build those integrations and have agents talk to each other with cords and all that.

Speaker 1: 34:48

Sometimes I hear a lot about A2A and MCP. Can you talk about the difference between those two and how it's different, sure?

Speaker 2: 34:57

Think of MCP, modelp, model context protocol, as a way to plug into other systems. They always say it's the usb of, yeah, yeah, usb. So you're taking and taking that agent and it's able to plug into other systems and and do things, so communicate with those systems. Um, you know, if you want to, to plug in to Azure DevOps, for instance, and create an issue or create something on the Kanban board, you can have it do that with MCP. Agent to agent is if you're building multiple agents that are acting autonomously. So you have one agent that might do one thing in terms of gathering data from documentation or reading documentation. You might have another agent that's acting on the behalf of the user to actually submit or even interact with the computer, so a computer used agent. So those two agents need to talk to each other and share information. That is what A2A, or agent to agent protocol, is doing. So it's communicating and able to allow communication between agents. Mcp allows an agent to plug into other systems and use those other systems.

Speaker 1: 36:18

So how would you know which is the best to use? Because at the moment, everyone is leaning towards MCP, but you can still use agents if that's your use case.

Speaker 2: 36:30

You're going to use both. So it's not a best, it's not an either or. So if you're building multi-agent systems, you're going to use A to A to have them talk to each other. If you're going to use external systems that the agents need to talk to, you're going to use MCP. So you're going to use both protocols, depending on how you engineer, architect your system.

Speaker 1: 36:52

Yeah, that's brilliant. So let's talk about a little bit of like some of the, because I remember you said you're a grandmaster, something like.

Speaker 2: 37:04

Yeah, there's another organization known as Grandmasters. Yeah, I work with them. That's the company that actually develops exams. So, yeah, I worked with the founder, lisa chrysler, for years. Yeah, uh, lisa's awesome and her organization, um, helps, you know, develop exams. Uh, and I've been a content development manager there, um, and you know, subject matter expert working on exams for many years, so I help her and her organization to develop exams and we've done exams all across the gamut in terms of technologies and areas.

Speaker 1: 37:47

But, yeah, Is it just tech exam or you do different type like Power BI, different kind of exams in Microsoft.

Speaker 2: 37:55

Different types of exams depending on technology area. So for me, as a content development manager, it'd be different types of exams. Yeah, it could be full gamut of in terms of technologies.

Speaker 1: 38:10

Yeah, okay, so you'll be like getting feedback from learners and how you can improve the contents, where that you can do it better, and whether it's do more of an illustration or more of a case studies.

Speaker 2: 38:25

Yeah, constantly updating the exams. So that exams are constantly updated, you know, making sure they're relevant, that they're testing the right things, that they're testing the right cognitive level. So you know, exams ought to be value add to the experience and advance your skills. They shouldn't just be show up and throw up. I would say in terms of memorizing things. It should be applied knowledge and not memorization. So a lot of the exams are actually what I'd say open book, depending on the exam, so you can actually, during the exam, access the Microsoft Docs site. So if you're taking a Microsoft exam, you can actually open up a tab and access the Microsoft Docs site as well.

Speaker 1: 39:13

Yeah, yeah. So I think one of the ways that you can do is making it more of a lab-based scenario than have a little bit of multiple choice, because multiple choice people can just assume it's 50-50. But I want to use more of a lab scenario, like to do things hands-on, than that. That's why I prefer to do that. That's why I prefer to do that.

Speaker 2: 39:34

Exactly, I completely agree with you. Yeah, the lab scenarios are the best and I love hands-on and applied knowledge that way. So I completely agree.

Speaker 1: 39:44

Yeah, it could be like whenever people complete applied skills, you're 50% ready for an exam because you got the practical experience from that.

Speaker 2: 39:56

Exactly, you learn a lot more being hands-on like that. I feel you know actually doing it, you learn, you know what not to do. Sometimes you know fail fast, as they say. But yeah, when you get hands-on sometimes it's you know you learn a lot more.

Speaker 1: 40:15

Yeah, so any last-minute advice for anyone that's new into. Ar or for developers that have not started their journey to use agents or MCP yet in their organization.

Speaker 2: 40:32

I would say I know the space is moving fast. It's a lot to learn. Things have changed even in the last six months to a year. Models are constantly evolving. I think we're getting the space together in terms of the protocols coming together, but I would say, from a learning perspective, leverage the resources I talked about. Get hands-on and experiment, try different things. You know, kick the tires on some of the frameworks that are out there and you're going to see a lot of advantages, you know, in leveraging AI at various areas. So the low-hanging fruit chatbots chatting with LLMs, retriever, augmented generation chatbots we can do those in 5-10 minutes. Like I said, we've got a GitHub library or GitHub project for that. It's the real power that comes into affecting business processes like supply chain inventory, some of the key concerns with multimodal that I mentioned. Get hands-on, but get hands-on, just get hands-on.

Speaker 1: 41:41

Plus, nowadays, microsoft makes it easier that you don't need to be an AI engineer to use AI. Anyone can use it. Even your family or anyone that's non-technical can easily use AI.

Speaker 2: 41:54

Exactly yeah, with Copilot, for sure. Yeah, M365 Copilot, you know leveraging that, just asking questions, Absolutely yeah.

Speaker 1: 42:03

No worries. So, as this episode is coming to an end, we always love to get to know our guests, so do you have any hobbies or things that you do in your spare time, brian?

Speaker 2: 42:16

I've been since I was a kid. I've been since I was a kid. I've been a gamer, a computer gamer, so I used to play competitive quake. Uh, so, in software was one of my favorites. John carmack, um, you know, was the founder of first person shooters. Um, yeah, he was my inspiration to become a computer programmer back in the day so, just just like, it's mostly online gaming, or is it like console, xbox, ps4?

Speaker 1: 42:44

Everything.

Speaker 2: 42:45

Yeah, everything. I mean obviously a little bit biased towards Xbox. Playstation has great exclusives. I'm a gamer through and through. You know I love my Switch, nintendo. You know Nintendo, nintendo's been fantastic in the gaming area, so I'm a gamer. I would say. I love video games. That's what really got me inspired into computers. In the first place, I wanted to build video games, but instead I built a line of business apps and now I build AI apps.

Speaker 1: 43:20

So, yeah, it's been a while. Okay, what makes you want to switch, then? Because before your vision went to play video games.

Speaker 2: 43:28

Yeah, so a lot of my friends got into it. I had a lot of friends that worked for Blizzard. A company actually got acquired. Blizzard acquired their company and it became Blizzard North and they built the Diablo series. So I had friends that worked in video games. They just worked and worked and worked. They worked so many hours they didn't have lives outside of their work.

Speaker 2: 43:54

I wanted to get married, have a family, I have two boys and do other things outside of their work. I wanted to get married, have a family, I have two boys and do other things outside of video games, and I didn't want to make it not fun anymore. I think if I built video games they wouldn't be as fun anymore. So I looked at that and that's how I pivoted into becoming a computer programmer.

Speaker 1: 44:17

To be honest, you make it sound like it's going to be more stressful. It's not working. Like balance and stuff. Yeah, you're constantly challenging each other about gaming, like stats and stuff. Yeah, it's okay to have your hobby, but there's a time you need to stop. So it's always good. Exactly, exactly. Yeah, because I kind of like playing Wii as well. So like's always good. Exactly, I kind of like playing Wii as well. So like Mario Kart and Switch and playing Call of Duty on the PS4 and stuff.

Speaker 2: 44:46

It's all good because you can have a gaming headset.

Speaker 1: 44:49

you can talk to people and just shoot people virtually. It's all good, it's all fun.

Speaker 2: 44:57

Yeah, a lot of fun.

Speaker 1: 44:58

Okay, so aside from your hobbies playing games, do you go to like much events like Microsoft events, community events?

Speaker 2: 45:11

I used to attend. I actually spoke at Ignite a couple different times. I do go to various events locally. We have a NET user group and Azure user group. I used to run with a friend of mine, matt, in Omaha. I used to run the NET user group in Omaha, omaha, nebraska. So I love community events, supporting community events. But yeah, I mean, I love all of that. I've spoke at various events as well.

Speaker 1: 45:44

Nice. So which event do you think is your best, or you don't have a best?

Speaker 2: 45:52

I would say Microsoft Build. I mean, if I had to pick.

Speaker 1: 45:56

Is it because more developer hands-on?

Speaker 2: 45:59

Yeah, more developer hands-on focus. You know more directly from the engineering groups and things like that. As a gamer, I'm going to be a little biased and I would say BlizzCon. So as a Blizzard gamer, I'd say BlizzCon out in Anaheim. That's my favorite convention or area as well.

Speaker 1: 46:20

Yeah, but I'm not sure. If you win you get like a free game or free council. Yeah that would be cool. Yeah, you could be winning like a world competition. Yeah, online gaming Okay, that's quite cool. I take it. Yeah, because I know I work at Microsoft Build and they have mostly open hack and database and stuff. So I think that might be more. But even if it's developer conference, but it more so hands-on than other other Microsoft conference like Microsoft Ignite which is coming on a few months' time.

Speaker 2: 46:59

Yeah, I think November correct. I have to look up the date. I'm trying to know that San Francisco?

Speaker 1: 47:07

Yes, correct. Are you going to be there? Are you going to be speaking? I'm going to try.

Speaker 2: 47:13

I'm going to say I'm going to try. I'm not going to promise anything, but I always try. It depends on you know, customer engagements and things of that nature, but I'm going to try.

Speaker 1: 47:24

Yeah, you can say you're getting, you meet customers, meet partners and get feedback and stuff. You're twisting, I don't know what to do.

Speaker 2: 47:35

So I probably will see you there if I get there, because in San Francisco, francisco it's quite a little bit more of a journey than Seattle, where I normally go there it's 11 hours from London, so it's not like a direct flight yeah, that's a long flight, though that has to be quite a long flight yeah, so I think the best yeah that's a long flight, though that has to be, you know, quite a long flight, yeah, so I think that's the best.

Speaker 1: 48:04

I guess we'll just have to either sleep on the flight or just take a stopover, Because stopover means you can rest a little bit and then continue the flight.

Speaker 2: 48:10

Yeah, yeah, yeah, it's even a long flight for me. You're coming from Tampa, florida, so I think it's a good six hour flight to the west if I did a uh, non-stop flight oh, whoa, so almost the same distance, but after distance wow, because you know the way it's quite far, okay, yeah, so how?

Speaker 1: 48:35

so let's think about a little bit about the community. So when did you get started with the community, your journey? So is it when you were young?

Speaker 2: 48:44

yeah, I mean, when I started out, uh, pretty much around when a dot net was just becoming a thing, uh, okay, you know, yeah, there was a gentleman named Philip Wolf that started the NET community in Omaha and the local meetups, and then Matt Rui and I took over that user group and that's really where I got into the user group scene and we also had the Heartland Developers Conference, hdc that we ran with a buddy of ours, joe, and we actually had Scott Guthrie show up.

Speaker 2: 49:23

So Scott actually presented in Omaha and we had several folks that would present at this conference as well. But we always participated in the local community. I participate here locally as well in the Tampa community. I have a friend of mine, joe, that also works at Microsoft. That was a big speaker in the community scene. He was with our dev advocacy group and would speak locally as well. So that's where I met Joe and would speak locally as well. So that's where I met Joe. So Joe and I would speak and handle those local community events as well, are you?

Speaker 1: 50:05

still organizing events in that user group.

Speaker 2: 50:10

No Other individuals have taken over the user group and continue to drive it. I don't want to say I've been too busy to help in local events. I've just been disconnected, I would say, from the events locally. I need to get back out in the community and helping out there. We also have a group at Microsoft that are the dev advocates that do a lot of those speaking engagements as well.

Speaker 1: 50:37

Even though, for example, you still get busy, you can still do it once or occasionally when you're free, just to help Once you're behind the scenes, because everyone, like I, know every situation. Everyone's different. You may get busy at times, but once in a while, just helping behind the scenes is's always helpful. Yeah, yeah, I agree absolutely, and then you'll be getting your community spirit back and as well. So that's that's pretty cool. Yeah, well, thanks a lot, brian, for coming on this episode. It's amazing to to know a bit about your journey into like ai and some of the scenarios that people can use ai ways as well. Yeah, thanks for having me, it's been great and also yeah, and anyone wants to know about gaming.

Speaker 1: 51:30

You can speak to brian. I hope I can hook you up. Yeah, pre-counel, maybe if you win yeah, yeah, there you go. I don't know about that, but yeah, okay, now all right, thank you, see you next time all right, thank you very much.

Speaker 2: 51:45

Thanks, nicholas.