WEBVTT 00:00.400 --> 00:01.720 So please humor me. 00:01.720 --> 00:05.560 I've got a couple more admin things before we get into action. 00:05.560 --> 00:10.640 So I wanted to talk about how I've positioned this course and whether it's right for you. 00:11.400 --> 00:15.720 You may be new to coding someone that's never written a line of code before. 00:16.040 --> 00:20.480 You might be someone that's already considers yourself an agent engineer. 00:20.920 --> 00:22.800 Those are the two extremes. 00:22.800 --> 00:24.200 Or you might be in the middle. 00:24.240 --> 00:29.160 You're already a Python coder or you are an AI engineer. 00:29.160 --> 00:34.000 Maybe you're even someone that's taken my LLM engineering course, in which case, thank you and welcome. 00:34.880 --> 00:42.240 But I'm here to tell you that this course is designed to appeal to everyone across that entire continuum. 00:42.240 --> 00:44.480 There is something here for everyone. 00:44.520 --> 00:47.040 The people in the yellow boxes in the middle. 00:47.080 --> 00:49.240 You're going to have the best time with this. 00:49.280 --> 00:52.920 It's most well positioned for you, that is for sure. 00:52.920 --> 00:54.280 That goes without saying. 00:54.600 --> 00:58.000 For the people that are new to coding, it's going to be challenging. 00:58.000 --> 01:02.830 There's no doubt you're going to have to have a lot of patience, but this course is okay. 01:02.870 --> 01:05.950 You'll be able to get there and build amazing things. 01:06.350 --> 01:10.630 For someone who's already an agent engineer that's built some agent platforms before. 01:10.630 --> 01:15.830 You may want to speed through some of this, but I've made sure that there's advanced material covered 01:15.830 --> 01:16.230 too. 01:16.270 --> 01:20.830 We're going to be doing some really interesting things at every point and building some great projects 01:20.830 --> 01:21.270 too. 01:21.590 --> 01:26.590 So I'm here to say, if you're new to coding, take some time with the guides that I've written. 01:26.590 --> 01:29.990 You'll find them in the guides folder when we install the repo. 01:30.150 --> 01:35.830 And I've used that as a way to build up your basic, your foundational expertise so that you can hit 01:35.830 --> 01:40.470 the ground running for Python coders and particularly for AI engineers. 01:40.510 --> 01:43.070 If you've taken the LLM engineering course, it's perfect. 01:43.270 --> 01:45.750 Then you're going to have a great time. 01:46.190 --> 01:49.150 And for agent engineer, focus on the projects. 01:49.150 --> 01:51.150 That's when we really put things into action. 01:51.190 --> 01:53.230 I'm really proud of the projects we've built. 01:53.230 --> 01:56.910 I think you're going to enjoy them, even if you've got some experience with this already. 01:57.830 --> 02:04.010 And for anyone that hasn't taken the LM engineering course, I'll be sure to put links to it in the 02:04.010 --> 02:05.690 course resources. 02:05.690 --> 02:12.130 And it's something that I really do see that as complementary, and it's not necessarily a prerequisite 02:12.130 --> 02:13.050 to this course. 02:13.050 --> 02:17.690 You can come into this course without having taken it for sure, but if you have taken it, you will 02:17.690 --> 02:24.130 be really well equipped for this course because it builds up a lot of depth of expertise in terms of 02:24.130 --> 02:29.770 what it means to work to to choose and apply LMS to solve problems. 02:30.290 --> 02:33.410 So in just a moment, we're going to be starting to code. 02:33.450 --> 02:34.570 I know you're antsy. 02:34.570 --> 02:35.610 I know you're waiting for this. 02:35.690 --> 02:37.370 Can't he stop talking and get to it? 02:37.410 --> 02:38.290 We will be. 02:38.410 --> 02:43.250 But I do want to tell you that the first thing we're going to have to do is set up your environment 02:43.250 --> 02:47.730 and setting up an environment, a big data science environment at the frontier of what's possible. 02:47.890 --> 02:50.130 It can be a bit of a painful process. 02:50.450 --> 02:51.450 I'm here to help. 02:51.450 --> 02:53.090 I'm here to make it go smoothly. 02:53.090 --> 02:58.320 But please do have something of a thick skin for knowing that you may hit some challenges. 02:58.320 --> 02:59.240 We'll figure them out. 02:59.240 --> 03:00.280 We'll get them fixed. 03:00.320 --> 03:02.920 We'll get you up and running the. 03:03.120 --> 03:09.080 This time, though, we're going to be helped out by a couple of really great friends in the form of 03:09.080 --> 03:12.640 the platform cursor, which we'll be using as our IDE. 03:12.840 --> 03:14.640 And I just love cursor. 03:14.680 --> 03:21.760 Cursor, of course, is the AI platform that's powered by Llms that allows us to be so productive, 03:22.040 --> 03:25.560 and it's built on VSCode and and it's just great. 03:25.720 --> 03:28.160 And so it's going to be a lot of fun working with that. 03:28.160 --> 03:34.120 But perhaps more specifically for building the environment, we're going to be using this product called 03:34.160 --> 03:40.120 UV, which is you can think of as something which is replacing Anaconda that I used in the LLM engineering 03:40.120 --> 03:40.680 course. 03:40.720 --> 03:44.680 And it's a sort of built on top of, of using virtual environments. 03:44.960 --> 03:50.080 It's really fast and it's really simple and it just works. 03:50.080 --> 03:52.000 I am such a fan of UV. 03:52.200 --> 03:58.030 It's actually a student on the LLM engineering course that that first drew my attention to it and said, 03:58.030 --> 03:59.350 this is what we should be using. 03:59.550 --> 04:01.830 And and it was too late for LM engineering. 04:01.830 --> 04:03.750 But it's not too late for this course. 04:03.750 --> 04:09.630 And it turns out that UV has become so popular that most of the agent frameworks that we'll be looking 04:09.630 --> 04:12.110 at already have UV at their core. 04:12.310 --> 04:16.190 Crew is, in a way, crew is kind of like built with UV, as you'll see. 04:16.350 --> 04:18.950 So UV is really easy to use. 04:18.950 --> 04:22.070 It is nothing like Anaconda for people that have done this before. 04:22.070 --> 04:26.710 And it's it's basically very, very similar to using virtual AMS and you are going to love it. 04:26.710 --> 04:30.470 So I know it's always annoying to have to bring in something new to the mix, but it will be worth it. 04:30.470 --> 04:31.510 You're going to thank me for this. 04:31.550 --> 04:35.710 You're gonna you're gonna love UV and it's written in rust and everyone loves things that are written 04:35.710 --> 04:36.390 in rust. 04:36.390 --> 04:37.230 So there you go. 04:37.230 --> 04:38.150 Get ready for that. 04:38.990 --> 04:43.830 There's actually one more difficult conversation that we have to have, and it's about APIs. 04:44.310 --> 04:51.270 So look, this course does involve making calls to frontier models like OpenAI and Deep Seek. 04:51.470 --> 04:53.510 And they come at a price. 04:53.510 --> 04:55.350 There is a price point there. 04:55.790 --> 04:59.980 And I want to make the point that it is completely up to you. 05:00.220 --> 05:01.020 You can choose. 05:01.020 --> 05:07.460 I use both open AI and deep seq, but you can choose to only use deep seq throughout for a much lower 05:07.460 --> 05:08.140 cost. 05:08.140 --> 05:09.740 And Gemini you can also use. 05:09.780 --> 05:12.060 And right now Gemini has a free tier. 05:12.060 --> 05:17.180 I don't know how long that will last, but so you can use Gemini for free for most of the course. 05:17.500 --> 05:24.460 You can also use Alama completely free just running open source models locally on your computer. 05:24.500 --> 05:25.540 No cost at all. 05:25.580 --> 05:27.460 But there are some trade offs here. 05:27.460 --> 05:31.220 If you want to use any kind of good model, it will take a very long time. 05:31.220 --> 05:37.540 If you use the quick models like llama 3.2, then it might have troubles being coherent, and you certainly 05:37.540 --> 05:42.740 won't get to see the kinds of results that I'll be getting when I'm using the frontier models, but 05:42.740 --> 05:44.220 that allows you to have a sort of balance. 05:44.220 --> 05:49.380 You can run things locally for free and see one set of results, and you can watch me do it with the 05:49.380 --> 05:50.140 frontier models. 05:50.140 --> 05:52.980 If you don't want to spend, the API charges yourself. 05:53.780 --> 05:55.380 So what are these API charges? 05:55.440 --> 05:57.800 Well, different countries might have different pricing. 05:57.800 --> 06:04.840 What I can tell you is that for me, running this entire course, I spent well under $5 on any of the 06:04.840 --> 06:06.680 LLM model costs. 06:06.800 --> 06:08.120 It's very cheap. 06:08.280 --> 06:13.360 OpenAI the first time you use it, you need to put down at least $5 upfront. 06:13.400 --> 06:15.840 And I kind of pay as you go that you spend against. 06:15.840 --> 06:16.560 So that's annoying. 06:16.560 --> 06:18.760 So you might have to do that if you haven't done it before. 06:19.080 --> 06:23.520 But other than that, the actual amount that we'll spend will be relatively small, a matter of a few 06:23.520 --> 06:24.200 dollars. 06:24.200 --> 06:27.760 And it's very hard to spend very much on deep sea because it's so cheap. 06:28.240 --> 06:35.440 Now, I will say that in the last week I have an option to use a proper real time market data provider 06:35.440 --> 06:37.800 that I use, and that cost me $20. 06:37.800 --> 06:38.880 But that's not necessary. 06:38.880 --> 06:43.480 There are free options as well, but if you get into it like I do then, then you might do that. 06:43.840 --> 06:45.760 And you know, I just want to make the point at the end. 06:45.800 --> 06:49.680 People do have like like people get frustrated about API costs. 06:49.720 --> 06:54.360 There's something about them that are a bit galling, because it feels like everyone is sort of nickel 06:54.470 --> 06:55.230 and diming you. 06:55.270 --> 06:58.390 You're getting charged a little bit here, a little bit there, another API cost. 06:58.430 --> 07:02.990 And I understand that even though I say these things are very cheap, a few dollars, it adds up. 07:02.990 --> 07:04.710 And that's not always cheap. 07:04.750 --> 07:09.270 And particularly with exchange rates and different economies, that might be quite a burden. 07:09.270 --> 07:13.830 But I do want to make the point that you have to keep in mind what's going on. 07:13.830 --> 07:20.470 When we run inference on these large llms, there are trillions of floating point calculations going 07:20.470 --> 07:20.910 on. 07:20.950 --> 07:23.110 This is heavy machinery. 07:23.110 --> 07:28.190 And so it's not like these companies are just just making extra cash off you. 07:28.590 --> 07:34.230 I'm sure they are quite profitable, but but most of it, the margins are relatively slim because there's 07:34.230 --> 07:38.550 a lot of compute costs associated with running these massive models. 07:38.550 --> 07:44.590 And if you think of the price of buying a new laptop or even a big like a gamer box that could actually 07:44.590 --> 07:48.830 run larger models itself, you'd be talking about thousands of dollars. 07:49.030 --> 07:55.300 So this kind of API pricing, while it can feel, as I say, a little bit galling sometimes having to 07:55.340 --> 07:57.220 pay each of these different APIs. 07:57.500 --> 08:01.420 It is worth keeping in mind the context of what's going on behind the scenes. 08:01.420 --> 08:03.620 These companies need to pay their electricity bills. 08:03.620 --> 08:08.100 There's a lot of compute, and we're getting real value from what we use. 08:08.100 --> 08:11.300 So I don't know if that helps or not, but it gives you some context. 08:11.340 --> 08:16.180 Please do keep in mind you don't need to spend anything on APIs if you don't want to. 08:17.620 --> 08:20.540 So there are three things for you to remember. 08:20.700 --> 08:26.100 The first of them is that there are resources that accompany this course, and they're really great. 08:26.220 --> 08:31.740 There's a website where I've put links, and I've got extra content in the form of YouTube videos that 08:31.740 --> 08:32.820 should should be helpful. 08:33.140 --> 08:38.820 There's in GitHub, there's a whole section with different guides to teach different kinds of skills 08:38.820 --> 08:39.700 that you might need. 08:39.740 --> 08:44.860 And there's some troubleshooting guides in there too that will fix problems for you and the labs. 08:44.860 --> 08:52.140 So the labs I'm updating them all the time, the labs you can think of them as like an ongoing a fluid 08:52.290 --> 08:53.010 resource. 08:53.010 --> 08:57.810 Every time you do a git pull to get the latest, you'll get refreshed with new content. 08:57.810 --> 09:03.690 Maybe not every time, but most times I'm trying to keep them to be a living, breathing resource for 09:03.690 --> 09:04.050 you. 09:04.850 --> 09:07.170 Then remember to stay patient. 09:07.490 --> 09:12.650 Remember that if you hit problems, if you hit roadblocks, that's actually a great learning opportunity. 09:12.650 --> 09:18.530 Because figuring out what's going on, figuring out how to solve this is one of the best ways to learn 09:18.530 --> 09:19.930 and try and go with it. 09:19.970 --> 09:21.970 Try and enjoy it as much as you can. 09:22.090 --> 09:26.090 There's some juicy projects and figuring out why they don't work first time. 09:26.090 --> 09:27.130 That's part of the fun. 09:27.530 --> 09:33.770 And then the third point is that if all else fails, or even if nothing else fails, then contact me. 09:33.770 --> 09:34.650 I'm here. 09:34.850 --> 09:35.850 I love this stuff. 09:35.850 --> 09:37.850 I'm actually super responsive. 09:38.090 --> 09:42.650 If you email me or if you LinkedIn with me, then I'll respond very quickly. 09:42.650 --> 09:45.690 Unless I'm in a meeting or I'm traveling or something, or I'm asleep. 09:45.970 --> 09:49.570 For people in different time zones, I do tend to reply really quickly. 09:49.570 --> 09:50.850 People are always surprised. 09:50.890 --> 09:53.350 In fact, a lot of people think that I'm I'm an AI agent. 09:53.350 --> 09:55.790 When I reply, they're like, do you have an agent of yourself? 09:55.790 --> 09:59.710 But no, it's the real me, I will reply, I love getting questions. 09:59.710 --> 10:03.590 I love answering so, so do feel free to reach out. 10:03.630 --> 10:04.630 Be in touch. 10:04.750 --> 10:08.510 I welcome ideas and thoughts and questions and whatever you got. 10:08.550 --> 10:14.870 And by all means, if you if you're up for it, then LinkedIn with me I some people uh, like to uh, 10:14.950 --> 10:18.310 don't feel comfortable doing that for some reason, but I'm very open to it. 10:18.310 --> 10:23.910 I love building a community of people in data science and that can support each other. 10:23.910 --> 10:28.350 That can if people are looking for for to hire someone or if they're looking for jobs. 10:28.390 --> 10:29.910 And I can help connect people. 10:29.910 --> 10:34.110 So it's a really great way for me to build and contribute to the community. 10:34.430 --> 10:35.790 And this is really important. 10:35.790 --> 10:41.430 But one of the things that people did from my last course is they would post projects that they've done 10:41.470 --> 10:46.350 and tag me on them, or things that they've done from the course, and I can then amplify it, because 10:46.350 --> 10:51.260 I can then jump in with some comments and, and make some observations or say how great it is, they've 10:51.260 --> 10:54.940 done something, and that will then be available to all the other students from the course. 10:54.940 --> 10:56.220 You can weigh in as well. 10:56.220 --> 11:00.020 And this has really happened, and it allows you to share things and amplify them. 11:00.020 --> 11:02.060 And LinkedIn is a great place to do that. 11:02.100 --> 11:07.060 And it will be something that that will be seen, perhaps by future clients of yours or perhaps by future 11:07.060 --> 11:08.300 employers of yours as well. 11:08.300 --> 11:09.500 So it's a really good thing to do. 11:09.500 --> 11:11.020 So I strongly encourage it. 11:11.060 --> 11:13.700 LinkedIn with me and share on LinkedIn. 11:13.740 --> 11:17.140 It's a great resource and I can't wait to see what you're doing. 11:17.500 --> 11:21.180 So at this point, you've put up with half an hour of me yammering away. 11:21.420 --> 11:23.780 Finally, it's time for action. 11:23.780 --> 11:25.020 We're going to the lab. 11:25.020 --> 11:26.780 We're going to set up your environment. 11:26.780 --> 11:29.140 The next video is for PC people. 11:29.140 --> 11:35.820 We're going to have a video for PC setup, and then the video after that is for Mac people and Linux. 11:35.820 --> 11:37.540 You should probably join in with the Mac people. 11:37.540 --> 11:39.260 It'll be similar enough for you. 11:39.380 --> 11:43.100 And then we will reconvene on the video after that. 11:43.100 --> 11:49.820 So it's going to be PC, followed by Mac, followed by us all back together again with fully built environments. 11:49.860 --> 11:50.700 Let's do it.