Introduction, exam guide and setup

0:00 hey this is Andrew Brown your favorite Cloud instructor from exam Pro bringing you another free Cloud certification course and this time it's the Azure AI engineer associate also known as the ai102 published here for free on free camp and the way we're going to achieve technical certification is by doing the lecture content Hands-On labs in our Azure accounts and as always I provide you a free practice exam so you can go sit that exam get that certification put in your resume or LinkedIn to go get that promotion or new job you've been looking to get if you love free Cloud certification courses just like this one the best way to support the production of more is to purchase the optional paid study materials that includes things like practice exams flashcards quiz lits um downable cheat sheets and more which you can find over uh at exampro and this specific one is exampro doai h-102 if you do not know who I am I've worked in the industry for I probably 20 years now and I've taught courses on adus Azure gcp kubernetes Oracle terraform Linux and more so you're in

1:11 really really really good hands but let's just jump into it [Music] ciao hey this is Andrew Brown and we at the start of Journey asking the most important questions first which is what is the Azure AI engineer associate so this certification also known as the AI 102 is an AI certification as the name implies focusing on two things the managed AI services and working with large language models okay so I just want to warn you that if you're taking a Microsoft Azure exam they are very code and script driven heavy you'll probably see code in your exam so uh developer knowledge is a must but you're in good hands because my background is a developer and we open as much code as we can to make sure that you are prepared uh consider this certification if you want to uh directly learn how to uh work with open AI large language models now there are other models available in the Azure AI uh studio and we actually do launch them just for fun but they're not on the exam and these are the ones they want you to know if you want to have deep knowledge on implementing aures managed offerings uh or you want to know how to work with rags and Vector

2:20 databases or if you're a cloud engineer uh upskilling to take on AI engineering responsibilities probably more focused on gen because the time of this video that's what people are interested in so let's take a look at our Azure road map and so this is kind of the recommended path now this is not all of the Azure search because I cannot fit them all on the screen but these are the ones that are in your uh perview so you should start with the a900 take that AI 900 and then move on to the AI 102 now the AI 900 covers both the certifications in lighter lighter formats whereas the AI engineer is focusing more on those managed AI Services luckily if you've taken my AI 900 you'll be very well prepared for the Azure AI engineer because I did so much more and Beyond programmatically here basically almost helping you pass the AI engineer at this

3:09 level but I've gone way more deeper in the AI engineer um with the manage services and a lot on uh gen because that is what people are really looking for um let's talk about how long it will take to pass this exam well if you're a beginner about 24 hours that means that you've never taken Azure exam before you never worked with llms or a Services you don't have experience with python maybe you have Cloud experience but not programmatically and uh not with Azure exams or LM AI Services probably longer than 24 hours could take you up to 40 hours to be honest but it just really depends on you for the experienced person they have taken the AI 900 they have worked with python code uh they have basic familiarity with LMS and AI man Services it'll be 12 hours if you took my AI 900 course you're probably looking at six hours because I really helped you out um but anyway you're looking about 14 14 hours average study time 50% lecture in Labs 50% practice exams I recommend spreading the the workload out across two weeks one to two hours a day take your time do not try to rush it in a couple days you'll not remember it um but that's that uh uh what does it take to pass exam watch the

4:18 lecture videos do the Hands-On Labs okay uh it's very very important because especially with Azure um documentation does not often match the um the apis and the apis do not match the console um Azure likes to move quick and break things and so if you really want to understand what's going on you have to do those Hands-On Labs even my labs might break even though I've done them four five six seven times the um Azure is weird where it's like you might think you have something perfectly built but it just decides not to work so um you know just be patient there and don't get stressed uh do paid on online practice exams to simulate the real exam and so you can go over to exam pro.

4:40 a102 to get those additional practice exams we give you one for free on the platform but you know buy those additional ones it's going to help you on the exam it's going to help me to make more certification courses for free for everybody uh let's take a look at the content outline so there are the following domains we have plan and manage an a solution Implement content moderation Solutions Implement computer vision Solutions Implement natural language processing Solutions Implement knowledge Mining and document intelligence Solutions Implement generative AI Solutions and remember that each domain has its own waiting this is determined with how many questions of domain notice that Azure has ranges for their freaking sections it's very frustrating you never know how many questions you're going to get on the exam I don't know why they do this I do not like that as a testing practice but that's just how Azure rolls uh where you can take this exam in an in-person test center or from online from the convenience of your own home Microsoft delivers exams via Pearson view online or the PE view network uh of test centers so there might be a school or private private um place that has a computer lab they might have a test center partner with Pearson view you go there you give them your ID and you use

6:13 their facilities to the exam if I had to choose I would always do it in a test center because it's so much less stressful and less things will go wrong these exams are proctored meaning someone is watching you as you do the exam to make sure you do not cheat so no business when you're doing these exams and uh this is not the correct graphic here I guess it's just kind of got filled in wrong but that's okay the passing grade here is 700 out of a th000 points so you need around 70% to pass aure uses scale scoring so just be aware of that um that even if you get 70% you could still fail so aim for 85% and that's what's going to help you out there are about 40 to 60 Questions on this exam that means you get 12 to 18 questions wrong there are no Penal for wrong questions you're looking at multiple choice multiple answer drag and drop yes and no and more so probably case studies and Azure just had so many different kinds of of question types um so just be prepared for that the duration of the exam is 100 minutes you get one minute per question no a bit more than that I would say like 1.5 minutes something like that so the exam

7:23 time is 100 minutes the C time is 130 minutes C time refers to the amount of time you should allocate for the the exam this includes the time to review the instructions Show online Pro uh Show online Proctor your workspace read and accept the NDA complete the exam provide feedback at the end uh this exam this certification is valid for 12 months so that means it has a one year CER uh onee period before you have to recertify that sounds intense except um Azure lets you do renewals for free okay so hopefully that gives you an idea of what's going on here we are going to look at the exam guide much closer in the next video um and you know I just want to point out that uh the way I I make my content it is I love to show you uh everything that includes mistakes so notice here that there was minor mistakes and I'm correcting them as we go uh this going to be the same thing when we do our lab content uh the reason why I do that is that um especially when working with Azure things never go as expected I could do the same lab three four five times and every single time I'm going to get a different result so it's very important that I show you the troubleshooting so that you get stuck you can do it too okay but um yeah I'll see you in the next video

8:34 [Music] ciao hey this is angrew brown and we are taking a look at the exam guide for the ai102 and so what I want to do is go through it and tell you where I stuck true to the exam guide and where I deviated as not all these services are worth your time but some of them are and there's some additional things that I added that probably will be in future uh exams but are really essential if you want to be an AI engineer so just understand that there going to be some differences I'm going to get my head out of the way so we can look at the PDF in all its Glory or sorry HTML page as we haven't had PDFs in many many years let's scroll on down here and so one thing they're going to note is uh need skills with python and C now we use Python uh for the majority of this

9:21 course and we do a really good job with it could you see C examples on the exam yes but if you know how to do in Python you can trans the skill's over so it's not going to be super hard if you do not know c um D below says you should know what rest is and we actually have to use the rest API in some cases where the sdks are in beta or they just don't work as expected um so we will use rest API and we use sdks and with me we always do as much programmatic stuff as we can because that's going to give you the best chance of being able to do it in the real world but also be able to remember it for the exam so I made sure that we do things programmatically happy quite a bit in this course let's take a look at at a glance so here we have plan and manage Azure AI Solutions Implement content

10:09 moderation Solutions Implement computer vision Solutions Implement NLP Solutions Implement knowledge mining document intelligence Solutions and Implement generative AI Solutions so let's take a look at the first one here which is plan and manage an AI solution so here they're just saying just know what all the Azure AI services are avail available okay and I'm just actually going to go pull up just for a moment here portal. azure.com and we're going to go over to Azure AI services and on the left hand side these are all the services that are available now not all these are used um so like down below here these ones have been reworked some of these just won't appear like personalizer Health insights we're not going to see that immersive reader we're not going to see that but there are ones like AI search computer vision face face face API custom Vision speech search language translator document intelligence also known as form recognizer but they just love renaming things here um so yeah we are going to cover the majority of them um and so you know I don't have a comparative section but we go through all of these services so by the end of the course you should know what to choose implicitly based on what we've learned um again get my head

11:23 out of the way here plan create deploy Azure AI services so responsible AI so I did some updates to responsible AI not much has changed in the last few years they have a couple extra documents but again not super hard to know what to do for that um here they're talking about Azure a services so how to deploy it and how would it work with a cicd pipeline implementing with containers so we show you both of those here we have diagnostic logging they talking about like managing keys and things like that this stuff is pretty straightforward so there's not a whole lot to really say about it um you know so we don't really cover these Part Parts in much but again it's just like most Services that's pretty part of the course I'm not sure why they have that text in there then we have Azure AI content safety now there is another API in point called um content moderation I can't remember if we use it in the course we definitely use content safety and content safety has content moderation in it but uh we might end up coming both apis but we definitely make sure we fully cover Azure AI content safety we have an images which is under computer vision

12:32 resource um it has its own SDK so here um we talk about generating image tags Azure AI Vision which is just Azure a AI Vision Studio we programmatically work with it quite a bit then down below we we there's for analyzing videos now we're on to NLP so natural language processing so we have Azure AI language uh so extract key phrases entities sentiment of text language used in text identify person personable identifiable information we do this quite a bit uh and one thing I want to point out is that we actually end up using in the course to set up our environment um Azure ml uh Studios um uh compute because this was the easiest place to do it so understand that we will be using Azure ml Studio not necessarily for deploying ml pipelines but at least for the environments and so we programmatically work through all these examples very thoroughly then we have Azure a speech these kind kind of kind of melt into each other because they're very similar services but we do do this I'm not sure we do ssml um yeah I think we I mean I remember doing it but the question did I

13:41 do in the a 900 and did I carry that content over I'm not sure um but uh that one's not too hard if it is missing then we have the translate language it's a bit confusing because some of these Services can do the same thing or they might be using utilizing the same SDK so I recall us translating but the question is did we specifically use the translate language resource I'm not sure then we have Implement and manage a language understanding model so this actually used to be called Lewis uh Luis but for some weird reason they decided to um get rid of that service it's still in in here down below here language understanding but now it's under the language service here but the experience is exactly the same okay the the difference is that they've added aure AI search so that you can bring in your knowledge base from there and um you can use open AI to help you generate things like utterances and so I just want to be very upfront with you that I do not go through and make labs for it again so I have a video that will show you how to get to it I have older videos on how to uh build language understanding projects

14:49 when it was called Lewis but the thing is it's just a name change and a shuffle and I just can't stand using the service because it is so clunky all right we have we have slides I'm not even sure if we shot the videos on it but we definitely have slides in the PDF you can download that will just show you screenshots of the updated experience but I'm going to tell you the old lecture content totally fine it will give you the contextual information that you want it's just a name change Lewis to language understanding and it's now found in a different location the same thing with this this is talking about Q&A this used to be called Q&A maker classic now it's just called custom Q&A moved under um the language service and I make it very clear how to find that in a lab that uh that's before the lecture and then the older lab content okay we have

15:34 document intelligence um so this I believe is form recognizer uh so they gave it a different name which is down below here uh so this is something that we did before and I believe I expanded on it in uh this course here by adding additional examples for Azure AI document intelligence but you will see that it is just form recognizer renamed we have Azure AI search which um before was called M cognitive search and it really was just a fulltech search engine but now it it is a vector database store and is very important when we're working with llms so I gave this one extra attention because it is such a useful service not if not if if you don't really like Azure search it's not that important but if you want to work with LMS professionally there's just a lot of stuff that you need to know and you're going to get full knowledge uh with the amount of work that I put into here then we have open AI service this one is about launching um open AI so if we go over to here I'm just going to quickly show you actually it's here on the left hand side we have Azure open Ai and so you launch this resource this allows you to then deploy large language models um but it does get a bit confusing because

16:43 there's a thing called Azure open AI studio and then there's Azure AI studio and they look really similar um and they do have some overlap but they are a bit different and I I make sure to spend a good amount of time making it clear uh the differen the studios and the services down below we have some generative stuff like J using Dolly uh using open AI models all sorts of stuff now something that is not in this exam guide that I definitely gave you because you absolutely should have it is promp flow so promp flow is a way of um uh you could say multimodal or you could say it's a way of orchestrating coordinating multiple actions with your agents and other llms and things like that but I thought it was super super important and even if it does not appear in your exam today there's no way they're not going to add this in the future because it is such a powerful service I think the reason why it's not in the exam guide right now is that it's in preview but it's so so thoroughly developed uh there's no way they're clawing this service back and so uh you're going to get that uh as an addition and you really should uh study

17:55 it um even if it's not on your exam okay but uh yeah so there you go um so the only thing again is I was a little bit lazy with Lewis language understanding and Q&A but you have the resources there from prior uh you'll just have to make that translation yourself okay I wanted to spend a lot of time with the llms and they don't really detail it uh very descriptively here but I I think there was fine tuning in here as well I didn't see it fine tuning yeah they kind of hide these as like on liners that was another thing that I really did a good job uh job was explaining fine tuning and showing you how to do fine tuning um but it's weird that they just have it as like a single line item even though it's so darn important um but yeah there you go okay [Music] ciao hey this is Andrew Brown I want to get you set up programmatically uh for how we're going to work with specifically AI services so when we're working with AI Services I like to work in jupyter notebooks because I think that has the most Synergy uh for development and so uh technically vs code can run notebooks and so we could do it locally we could do in GitHub code spaces but where I'm going to do it is I'm going to

19:08 do it uh in a notebook within Azure machine Learning Studio now Azure AI Studio which right now is still in preview has the ability to launch um notebooks however it's just utilizing Azure machine learning uh notebooks underneath and it actually makes it more confusing using a studio because at this point it's just not uh well-developed service um so I'd rather just save you the pain and we're going to go ahead over to Azure machine learning um and what we'll do is spin up one here and from there we will then um have our compute there all examples are going to up in the Azure examples repo um I just started this so depending on where you come in uh there could be lots here or little um but at this point this is my starting point here I'm going to go ahead here and create a new work space so we'll go ahead and do that um I'm going to go ahead here just say um AZ or just say AI workspace could be AI or ml workspace doesn't really matter so I'm just going to call this my AI workspace uh it could matter where we

20:17 launch this I'm going to launch this in um maybe West us the reason why is that uh when we're dealing with um machine learning Services some types of compute or gpus are only available in particular areas uh I know like when you're doing fine-tuning uh uh at the time of this video they only appear in West us and um Sweden so I'm going to stick with West us it's not normally where I would put things but that's where we going to put things here today it's going to create a storage account it's going to create a key Vault it's going to create application insights uh which I'm not sure I guess we do need it we can't say no um and container registry which I'm just going to leave blank for networking options we have public private uh private with approved outbound pretty standard uh for um Azure services to provide three levels of network

21:01 isolation but this UI looks a little bit different than some other places but we'll go ahead and hit review and create and so we will go and create um this Azure machine learning workspace okay so I'll just pause here we be back in just a moment all right so that was just the review stage we'll go ahead and now create I thought it was creating sometimes I think that it's creating when it's not um but it has to do with that combination there so now it's deploying uh and we will wait for that to uh complete its deployment okay all right so that resource is completed let's go ahead and um launch this here so I'm going to now go ahead and launch studio now launching Studio doesn't cost anything so right now we aren't really worried about any spend with the exception of storage accounts and anything else that's collected there apparently there's a lot of preview features uh that's cool wow there is a lot of preview features um and uh that is very interesting looks like there's even stuff for Visual Studio code but again I want to keep it simple and I want to work um with an environment that uh is going to be easy so over here we have environments and we have compute I'm trying to remember which one it is

22:14 um no these are predefined environments so this is not exactly what I'm looking for but you know I want to just point out that when we start working with Azure AI services or open AI Services things might appear here as well so um just always double check this environment uh because it can get overlooked but I think what we're looking for down below here is compute and so here it says get started with Azure machine learning notebooks and R scripts by creating your compute instance and so that's what I want to do um is I want to create a compute over here so what is the cost for compute here I do not remember but they do have servus instances which is really interesting if we go over here man servess instances utilized by prompt flow compute session um um right now I'm not super confident with that I'd rather just stick with regular compute instances because that stuff is very clear um but I can to see if I can look up pricing very quickly here so we understand what we're getting into now we actually might you know what I'm going to do I'm just going to start clicking here because it might just tell us the cost oh there we go okay you know sorry I'm just used to using things like AWS where they do not tell you um so since the last time I use this this is different which is fine but notice we

23:25 have some Associated cost so we have uh 8 16 cents an hour 28 cents an hour 37 cents an hour so it really depends on what you're doing so if you are um you know working with something you might have to have something and then spin it down so you'll have to uh be mindful of how long you have this running for but if we're doing AI it's very hard to avoid any kind of spend not at the top we have CPU and GPU if we're using the Azure AI Services API we aren't doing the compute the who's doing the compute is going to be um it will actually be Azure itself right so we might come back here and spin up something that has computer attached to it but for now all I really want to do is I just want to uh work with CPU for general purpose the lowest possible cost here it looks like this is really low looks like there's also some other options if we can go cheaper I would love to do that because again we're going to just use apis for now there we go and so we wouldn't obviously want to do that uh for anything else but when we're using apis but this is what I'm going to spin up

24:30 here today we'll go ahead and hit next um oh so it can shut down over time which is good I'm going to just put in 30 minutes here just in case so that's really nice um do we want any other options here so configure security sets such as SSH virtual Network manage identity for compute really depends on what we're doing right now again I'm just trying to show you that experience of spinning up uh one of these compute instances um but I don't think that we need to do anything here create a computer instance on behalf of another user well we don't need it on behalf of another user it's going to be us do we need to assign any identities we might want to do that but for now I'm going to go ahead and just hit next um I don't need any startup scripts and here we can add environments so I guess if we create an environment ahead of time we could attach one I don't need any of these so I'm just going to ignore them so go ahead and hit next next we'll leave tags alone and we'll just review what we have so we've chosen the cheapest possible option um the compute name is called Andrew 2 not the best name I don't like that name I'm going to go ahead and just say AI compute since this resource can shut

25:44 down it makes me think that uh we can shut it down when we're not using it which is great um we'll just say AI compute uh AI AP uh managed API compute okay and uh that way we know exactly what this is for don't really need single sign on but that is totally fine we'll go next um I'm not going to add environment because I'm not sure what to do for that right now and we're going to go ahead and hit create and so that is going to um create our compute if you have a larger type of computer with gpus attached then it might take longer to spin up this should not take to too long so I'm going to pause pause here I'll be back here in just a moment okay all right so now our uh compute resource is running and we have a few options the vs code web is um the newer option um but I'm going to stick with Jupiter lab uh I'm going to assume Jupiter is just Jupiter notebook but Jupiter lab is the more modern version of Jupiter notebook so we'll go ahead and open this up here again this whole video is just to um spin up a resource so here it says

26:55 jupyter lab is suggested and so it's just is telling us what is uh being included in here uh and so what we're going to need is we're going to need some type of Kernel um and so usually kernel appears up here but I don't see it so I'm trying to remember how we specify our kernel kernel usage is missing H so it's possible that we should have uh included one initially here but we do have some notebooks uh like azure uh ML and things like that um so let's try 3.10 SDK version 2 and so I'm hoping if if I do this it's going to add a kernel for us let just say print hello world and we'll run this here and it runs it so yeah now we're getting kernel usage which is great um and so that's that's what we wanted here okay so now that we have that we are in good shape and um I mean what we should really do is just make sure that we have

28:07 API access um the way we can do that is just by um testing some kind of API so just give me a moment to think of some kind of like test case that we can do here okay all right so I think what I want to do is I want to go ahead and again use some kind of AI service again we'll probably do this at some other point but I just want to have some kind of use case that we can test here um I think that working with text is always