Microsoft Community Insights
Welcome to the Microsoft Community Insights Podcast, where we explore the world of Microsoft Technologies. Interview experts in the field to share insights, stories, and experiences in the cloud.
if you would like to watch the video version you can watch it on YouTube below
https://youtube.com/playlist?list=PLHohm6w4Gzi6KH8FqhIaUN-dbqAPT2wCX&si=BFaJa4LuAsPa2bfH
Hope you enjoy it
Microsoft Community Insights
Episode 17 - Innovate Azure with Pulumi with Alec Harrison
Discover why Pulumi could be your new go-to tool for infrastructure as code in our latest episode with Alec Harrison, a Microsoft MVP in AI and Azure. Alec shares compelling reasons to choose Pulumi over other well-known tools like Terraform and Bicep, particularly in light of recent shifts in Terraform’s terms of service and IBM’s acquisition. Learn how Pulumi’s support for multiple programming languages and cloud providers, including AWS, Azure, and GCP, can streamline your multi-cloud deployments with unified scripts and efficient processes.
Welcome to Microsoft Community Podcast, where we share insights from community experts. You stay up to date in Microsoft. I'm Nicholas and I'll be your host today. In this podcast we'll dive into Pulumi and Azure. But before we get started, I would remind you to follow us on social media so you never miss an episode, so you help us reach more amazing people like yourself. Today we have a special guest called Alec Harrison. Could you please introduce yourself please?
Speaker 2:Yeah, sure, I'm Alec Harrison. I am a Microsoft MVP in AI and Azure. I'm located in the US, in Omaha, nebraska. Thanks for having me.
Speaker 1:Thanks, Welcome. So since our session is called Pulumi, could you give us some like explain what's Pulumi and how is it used?
Speaker 2:Yeah for sure. So Pulumi is an alternative for infrastructure as code in the cloud. The most applicable thing it would be equivalent to or similar to is probably Terraform. I think a lot of people have heard and used that product. Pulumi is kind of similar to that, except it allows you to write your infrastructure as code in the language of which you choose. Except it allows you to write your infrastructure's code in the language of which you choose. So think of like AWS has their CDK where you can do like TypeScript, all sorts of languages. Pulumi does that as well, but it's cloud agnostic.
Speaker 1:So whether you're in AWS, azure, gcp, you can go ahead and write your Pulumi file to go to any cloud of your choice. Okay, so why would someone choose to use Glimmy in their environment and stuff like Terraform and Bicep?
Speaker 2:So I think Terraform scared a lot of people away when they first made their biggest change in their terms of service and the open tofu fork happened. So that was when they changed their terms of service. The billing model maybe looked like it could potentially change. The community decided, hey, we're going to fork the open source repo and we'll manually keep that up to date. That's a lot of overhead and a lot of pain, particularly like somebody has to go out and do that and keep it up to date While the tech is being widely used. That's not a big issue, but you're then not getting first-party support right. You're getting the somebody's taking the fork and updating it manually. The other thing is HashiCorp, or Terraform, was recently purchased by IBM. I don't think there's been any major changes right now but, as far as I can tell, like IBM owns them now, they could totally change how the system works or how the technology works. So I think that's another reason why I'd potentially choose it Instead of using, like your primary cloud providers, ones like you have, Bicep Arm, CDK CloudFormation Instead of that, if you're running multi-cloud or want the potential of you know having a single source of truth while, like you, still have to write basically an if statement to say, if Azure do this, else do this for AWS, you can still write it in one language or one technology suite.
Speaker 2:Right, so you could share scripts, you can share them around, and that would be for. If you don't want to be just in the proprietary languages, so say I don't know, maybe my product has a requirement. I want to be multi cloud. If I'm trying to use the first party support tools like CDK and bicep, I would have probably two separate scripts that have two separate deployment processes that you might be able to have status checks in your pipeline, but it's not going to be the same deployment process to both, right? So I would have to say, hey, go deploy to AWS, go deploy to Azure, where, if you do it in a tool like Pulumi, Terraform is similar too. You can say, hey, here's my script for both clouds, go deploy it. It's the same deployment process and you could roll back as you see fit. Pulumi and Terraform 2 also both have state files, so you can kind of see what the state is of your cloud and hopefully, try to get a little bit more monitor drift as well.
Speaker 2:So if your resources are drifting you can kind of get some alerts with that as well.
Speaker 1:Okay, because I know that Terraform have their own SDK as well than Plumi.
Speaker 2:Yeah, they have their own SDK and CLI that you can roll in and use. Plumi also has Plumi AI, which is new. So let's.
Speaker 1:So let's talk about a little bit of AI at the moment. So how are you using Gen AI and AI overall? Are you using it at work or in your spare time?
Speaker 2:Yeah, so I'm actually helping a company right now. It's about a $2 billion manufacturing company and we are working on building out some gen ai use cases for them. So we have one that we're hoping to go to production here within the month, which is kind of helping some end users with their questions. I don't know if I can demo that specifically, but what it does essentially is somebody can ask a question and then it pings all of their internal documentation and tries to answer it because they also have a traditional call center. So what it's trying to do is make the first person that, um, that bot, essentially right that hey, here's the answer quickly, and then we're also providing citations. So the experience is very similar to I don't know if you've used google lately um, how to use github copilot, let's do that.
Speaker 2:Okay, that's not going to do it. Um, what I'll try to show you is the gemini model from sometimes we'll pop up here and try to answer your question. It's very similar experience to that that we've built, where we'll say, hey, here's what we think the answer is, here's the citations that we think go along with that, so that they then can click links like you would in Google. So maybe they don't trust our AI answer. They can go ahead and click here and see more data, whether it's their own internal, you know, pdfs, their own internal documents, their own internal, what have you? Okay, so it's only for internal use.
Speaker 1:Is it only for internal use, like document search and stuff? Okay, Is it like a? Is it on production? Or is it just normally like a proof of concept, Like is it on production?
Speaker 2:or is it just normally like a proof of concept? We're in the process of deploying it to a small subset of users. So I think when you talk Gen AI, there's two big fears right now it's security and it's cost. I think, security-wise, the company I'm at is fairly confident that they feel like security is okay because we're using all Azure products. So Azure, you know, keeps all of your data within your specific subscription. They don't use it to go back to the models, to train it, it just stays where it is right. So, security-wise, they feel really confident that Microsoft and Azure has their back there.
Speaker 2:The biggest thing they're concerned about is cost. So I think we've proven with different POCs that we think it's accurate enough to be put in front of end users Because, worst case scenario, they still call the call center. So we still have that backup, that escape route, if you will. And then, best case scenario, they get their answer and they're off. So that leaves cost and right now we just don't know, kind of, how well people will use it or how much people use it. One thing I will call out is, with Azure open AI, gpt 4.0 mini, azure OpenAI, gpt 4.0 Mini the cost for AI has pretty drastically dropped.
Speaker 1:So this is per thousand tokens, but I guess you can just yeah, but I guess you can just customize it and what you want. But you just have to keep testing it like whatever you want, search and document only, and limit it.
Speaker 2:Yeah, and we've locked it down too. There's a configuration in the API to say, basically, only use my data I've provided you. So that should help cut down on hallucinations, hopefully, because then you can't you know, for example go find an answer on the internet that's completely wrong or off-base or not applicable to our company. So that's how we kind of solve that one. Just say you don't know. If you don't know Otherwise, only use the documents we have.
Speaker 1:Okay, that's good. So, aside from the application that you build or you're going to do from work, what other exciting things you've seen around Gen AI that interest you at the moment? Yeah, I think, Whether it's AI, Mac OS or any Go AR open.
Speaker 2:any Copilot Studio is a really powerful one for non-developers. So if you're not a traditional app dev, it gets you, I would say, especially like chat with your data right, being able to connect to external data sources and drop it into different things. They have a bunch of integrators out there, like Teams, slack, any of those. It's really exciting because then people with low-code, no-code skill sets can jump in and build these chatbots that could potentially help their companies out.
Speaker 1:Okay, yeah, that's good so in your opinion. So I just want to ask what do you think AI would help you to do in the future? Is that a general opinion? Yeah, I think Over time.
Speaker 2:I think it'll help people be more efficient in the roles they're doing. I don't think we're at the point where it's really going to replace people. I think it's just going to help people be more efficient and then hopefully then move people closer to the problems they're solving Right. Like, instead of writing as much boilerplate, I can ask GitHub copilot to do that for me. Instead of spending time creating a PowerPoint, I can ask Office 365 Copilot to create a pretty PowerPoint for me and then I can focus on the job I was brought in to do. Right, I can help write code, I can help solve business problems, and then the things that are maybe less important but still time consuming, I can get an assistant of ai to help me, you know, create it and get going there yeah, is there any particular trend?
Speaker 1:do you see that's coming up in the field of ai or in general?
Speaker 2:I think some people got really excited, and I feel like AI kind of got overhyped. I think we saw this in a lot of other fields too, like the internet and other technology. I think blockchain had a huge hype wheel too. So did especially with like NFTs, or non-fungible tokens, when they were like selling art for crazy amounts of money. Uh, bitcoin's another one.
Speaker 2:Like a lot of these technologies, have super big hype cycles yeah I feel like we're waiting for the next big announcement from a big llm company, because I feel, uh, if you look, I think nvidia's stock is down or dropped a ton lately. I think the hype is kind of subsiding and companies are now taking a more critical look and actually investigating of like will AI drastically change what I do? I don't know say like 5% or 10% more productive, but it's not going to make it. So you're now, you know, a one-man company making $3 trillion. It's not anywhere near there.
Speaker 1:Yeah, it's not going to replace us humans. It's still going to be there in a way.
Speaker 2:Yeah, I think the phrase I keep hearing is you won't be replaced by AI. You might get replaced by a human using AI.
Speaker 1:So this episode is coming to a close, so what other advice would you give people when starting with Pulumi or AI in general?
Speaker 2:With Pulumi. I would say I think both Just start building. You don't have to fill out a thing for Microsoft anymore to actually get access to Azure OpenAI, so just start building. See what it costs to ask a question. Learn how to index your data. There's a lot of quick starts from Microsoft on how to get going specifically with chatting with your data. There's a lot of quick starts from Microsoft on how to get going specifically with chatting with your data. There's a lot of services in Azure to make that a super slick and seamless process. So I'd say, get started. Same with Pulumi Start building. If you want to learn about DevOps, if you want to learn about pipelines, publish your own website. I have a silly blog that was a Gatsby blog. That's how I got started into more infrastructure as code and pipelines. I just started building something and then go from there. Find the pieces you like, find the pieces you don't, and then, yeah, that's how you learn other concepts and other things. I'm a very practical person.
Speaker 1:I'd rather learn with my hands dirty than from a textbook, if I can help it. Yeah, yeah, it's all by learning.
Speaker 2:By doing so is there any other advice you would like to give to anyone? One thing that's really good. If you're looking for like an artificial deadline, I would say try to get a talk accepted at a user group. So try to do something similar to this write a 30 minute hour, talk about some concept and then go learn it. I guarantee you some people might know it better than you, but for the most part, communities like that are very nice and helpful and they'll help you out, or you can present it to them and learn some new stuff and network, and then they can. If you hit any roadblocks, just say like I couldn't figure out x, and somebody in the crowd might be like have you tried this? Or I had the same issue. Here's how I solved it.
Speaker 1:So just go and share what you're learning. Speaking of community, you're quite involved in the community, alec, so can you tell us any?
Speaker 2:of the community involvement. Yeah, I host a few user groups here in the States. We are doing a AI workshop October 17th. It's going to kick off 3 pm Central time in the US, but it's hybrid, so if you search Iowa Microsoft Azure User Group on Meetup or follow me on LinkedIn, I'll be posting about it. I also run a weekly podcast that you've been on, called Azure Cloud Talks, where we talk to other MVPs, other people in the community, about technology things, about how they got into their career and just anything that might be irking them at any point in time, just to show you know we all get frustrated, we all get past it. It's just part of life, part of growing.
Speaker 1:Yeah, Okay. So before we close the episode, we always like to know a bit more about the yeah yourself. So are you going to any? Are you doing any events or going to any conference?
Speaker 2:uh, I don't think I have anything right now. It's kind of a talk preparation season so I'm writing some you know submissions for next year's conference, speaking. But yeah, nothing crazy in the books right now, just looking forward to next year and seeing the next conference season spin up.
Speaker 1:Okay, Any any particular conference that you're looking forward to next year?
Speaker 2:I always have a good time at KCDC, so the Kansas City Developers Conference, that one's always a blast. Microsoft Build has been quite a bit of fun. The last couple of years I've gone to Trying to think of what else. Heartland Developers Conference here in Omaha is a thing that's been going on for a while. It's fun. It's just down the road for me, so I don't have to travel very far for that one, which is nice. Yeah, those are all fun conferences that I enjoy.
Speaker 1:Yeah, I heard that KCDC is quite popular developer conference as well. It's quite hard to get into as well. You have to know someone, don't you Know, someone quite hard to get into as well. You have to know someone doesn't know someone.
Speaker 2:I forget what it was, but I think there was one track. I think it was the non-technical track. You statistically was about the same as I forget what it was. It was some crazy stat. You're more likely to be struck by lightning to get in or something like that, because they have an insane amount of submissions and they took like 30 of them. It's a little bit easier if you do a technical track, but not much.
Speaker 1:You probably got more chance when you do like a AMA or a group chat about something.
Speaker 2:Yeah, I know my co-host for my podcast. Yeah, brian Gorman seems to always get one in, so he's a good person to emulate. He's an MCT and MVP and runs my podcast with me. I think the last two years he's gotten in.
Speaker 1:Nice, okay now I have podcast with me, he. I think the last two years he's gotten in okay, nice, okay. Now, as of thanks, uh, where can people get into, where can people get in touch with you? The best place if you want to advise anything yeah, I'm pretty active on linkedin.
Speaker 2:uh, if you just go to linkedincom slash, I think I'm at Alec13355. So A-L-E-C-13355. I'm there. I have my LinkedIn learning course there. If you want to learn a little bit more about AI, or if you just want to follow anything else I'm doing, I typically post on there.
Speaker 1:Okay, what course is that? Is that the AI course, or?
Speaker 2:okay, what course is that? Is that the ai course? Or, yeah, it's a course over gen ai. So we cover specifically the ai 102 exam, the gen ai portion. So if you want to learn more about ai, you can sit learn a little bit more from me. And uh, yeah, it's on LinkedIn Learning.
Speaker 1:Okay, yeah, no worries. Thanks for coming to this episode, alec. In a few weeks it's going to be on social media. Okay, thank you.
Speaker 2:Yeah, thanks for having me.