Description Transcript
Ever wondered how AI stacks compare to your regular tech stack? Developer Advocate Angelik Laboy Torres breaks down the differences, explores the overlap, and helps you understand how these stacks power modern software. Whether you're building AI applications or curious about how AI fits into your existing developer workflow, this breakdown will help you understand the evolving software architecture behind AI development.
🔖 Chapters
00:00 - Intro
00:36 - Tech Stack
00:57 - FrontEnd
01:06 - BackEnd
01:16 - Database
01:22 - Infrastructure
01:32 - AI Stack
01:49 - Data Layer
02:04 - Compute Layer
02:17 - Model Layer
02:33 - App Layer
03:03 - Observability
03:14 - Managed LLMs
03:25 - UI
03:33 - Frameworks
#dev #ai #tech
Read more 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!
