0:19 Who would have thought it would get 0:21 so busy... 0:21 Though I wonder 0:23 if we can accomplish things that we haven't 0:25 thought as possible 0:27 before. 0:28 Okay, yeah I'm done! 0:30 Why haven't we said what this video is all about? 0:32 It's all about the STACK! 0:36 Most of us have heard or have a tech stack. 0:38 A bunch of tools that allow us 0:40 to build and run an app. 0:41 Now, it can include a flavor of programming languages, 0:45 infrastructure, and frameworks. 0:47 Much like a recipe at a restaurant 0:49 and like we need ingredients, 0:51 we also need the people to staff the location. 0:57 Let's say we're at a coffee shop! 0:58 A customer will first go to the customer area 1:01 and interact with me the barista. 1:03 As the first line of defense, 1:04 I'm known as the Frontend. 1:06 I then place my order with the version of me doing the coffee. 1:10 I need to process what the customer ordered 1:12 known as the Backend, 1:13 because the machines are always
in the back. 1:16 The backend needs to locate each ingredient 1:18 so we visit the storage room, 1:20 aka the Database. 1:21 With the coffee beans and milk in hand, 1:24 I have to check if I have sufficient energy to run the shop. 1:27 I then look up to the sky 1:28 and the cloud delivers me the 'okay'! 1:31 So, let's continue, let's rework 1:33 the same exercise but with AI stacks in mind. 1:39 AI stack is a tech stack but for AI models 1:41 and it needs tools, frameworks, and infrastructure 1:44 but this will enable training, 1:46 deploying, 1:47 and running AI systems. 1:49 Now, the names for each layer will change, 1:51 but it will remain the same process 1:53 In this scenario: 1:54 there is a Data Layer that has the ability 1:56 to learn from the past customer orders, 1:59 making sure 1:59 to always collect user data. 2:03 The order is then handed off to the Backend 2:05 which is now a Compute Layer 2:07 capable of high processing power 2:09 to make coffee quicker. 2:11 Here the engine is running, 2:13 coffee is made, 2:14 orders are processed, 2:15 and everything runs efficiently. 2:17 Paired with a Model Layer, 2:19 predictions can be made for the customer 2:20 so it doesn't have to go 2:22 and select an ingredient that they don't like 2:24 but instead, give the best
recommendation that fits them. 2:27 I can then be fancy 2:29 and go ahead and recommend a new drink 2:31 that I know that they will like 2:32 and, before we even finish, 2:34 we need a way for the customer to talk to us. 2:37 Think of the application 2:38 as a type of Frontend 2:39 since it interacts with the user 2:41 but it's not always UI based. 2:43 It can also be API based 2:45 or just in the background. 2:47 For us, in this example, 2:48 a self-serve kiosk can be present to recommend 2:51 drinks based on the weather 2:52 or a drive-thru with voice assistant, 2:55 similar to an API sending orders, 2:58 or a recommendation menu based on sales trends, 3:02 think, Netflix. 3:03 In this process, 3:04 there's always somebody watching, 3:06 monitoring, 3:08 and judging 3:09 to see if all AI models are accurate, 3:12 reliable, 3:12 and able to scale. 3:14 Other concepts you might see lying around 3:16 are the Managed LLMs 3:18 which are Large Language Models 3:20 hosted on the cloud 3:21 that you don't need to train on your own 3:22 and that's part of the Compute Layer. 3:24 Then, you have the UI 3:26 which are the user interface 3:27 or how you interact with the AI. 3:30 That's part of the Application Layer. 3:32 And, then lastly, 3:32 you have Frameworks 3:33 which are pre-built set of tools and libraries 3:36 that allow for AI models to be trained, 3:39 deployed, 3:40 and built! 3:41 That also includes a high-level abstraction, 3:44 pre-built functions, 3:45 and an evaluation of performance. 3:47 And, that is part of the Model Layer 3:49 and, so with that, 3:50 I leave you and I hope to see you next time!
