WEBVTT 00:00.240 --> 00:04.400 So, as promised, I'm coming back now to tell you what this course is actually about. 00:04.720 --> 00:06.600 We're going to be doing three things. 00:06.600 --> 00:11.800 We're going to be working on theory, talking about what agents are and what an agentic architecture 00:11.800 --> 00:12.600 looks like. 00:12.640 --> 00:14.440 We're going to be talking about frameworks. 00:14.440 --> 00:19.640 So working with actual platforms that allow us to build and deploy AI agents. 00:19.960 --> 00:25.680 And of course, we're going to be getting hands on rolling our sleeves up, building some projects. 00:25.680 --> 00:29.080 Some of the projects will be more on the sort of entertaining side. 00:29.360 --> 00:34.000 Some of them will be much more on the commercial side, building real functionality. 00:34.280 --> 00:35.720 All of them will be fun. 00:36.200 --> 00:42.960 So I've organized the curriculum into six modules that I call the weeks week one through to week six. 00:42.960 --> 00:47.480 And I do that because I want to give you some motivation to be pushing on through this and keep up the 00:47.480 --> 00:48.040 drive. 00:48.080 --> 00:51.840 But you can take it at whatever pace you like, and if you can do it faster than six weeks, then so 00:51.840 --> 00:52.680 much the better. 00:52.840 --> 00:54.320 But, but that's how I've structured it. 00:54.320 --> 00:56.320 And each week builds on the last. 00:56.680 --> 01:00.120 The first week that we're on right now is called foundations. 01:00.120 --> 01:02.080 And it's all about grounding. 01:02.080 --> 01:04.360 You build building out the basics. 01:04.400 --> 01:09.560 Talking about what it takes to to have an agentic architecture, how different Llms can interact. 01:09.560 --> 01:15.200 And we're going to be building an agentic solution just with Llms interacting natively, not going through 01:15.200 --> 01:16.040 a framework. 01:16.320 --> 01:23.400 But in week two, we introduce our first framework, the OpenAI agents SDK, which is a beautiful framework 01:23.400 --> 01:28.040 because it's so simple and elegant and flexible and and will build some things with it. 01:28.080 --> 01:29.440 We'll add guardrails. 01:29.440 --> 01:30.960 It's going to be really great. 01:31.520 --> 01:35.520 In week three, though, we move to the very popular crew. 01:35.560 --> 01:38.520 I I'd say I'd say that crew is the fan favorite. 01:38.520 --> 01:39.800 People love crew. 01:40.000 --> 01:43.000 It's it's on the sort of low code end of the spectrum. 01:43.040 --> 01:47.920 It allows you to to write some configuration and use that to define your, your crew of agents that 01:47.920 --> 01:49.000 will solve your task. 01:49.160 --> 01:55.160 Um, and it's and it's a lot of fun to work with, perhaps on the opposite end of the scale in terms 01:55.160 --> 02:02.040 of, of of code, of low code to to to full code is Landgraf, which is very sophisticated, quite complex, 02:02.080 --> 02:03.360 very powerful. 02:03.360 --> 02:05.440 And we will be using all of its power. 02:05.920 --> 02:09.910 And in week five another powerful one is Microsoft's Autogen. 02:10.070 --> 02:13.070 Autogen is actually a couple of different things, and we'll look at all of them. 02:13.470 --> 02:19.110 But one of the things it is, is like an environment where agents can collaborate remotely. 02:19.150 --> 02:20.310 And that's going to be fun. 02:20.790 --> 02:28.990 And we wrap it up in week six with our capstone project and introducing MCP, the Model Context protocol 02:28.990 --> 02:30.070 from anthropic. 02:30.230 --> 02:31.750 Incredibly exciting. 02:31.910 --> 02:34.190 And it's really taking off in a big way. 02:34.390 --> 02:41.750 This is an open source way that different models can connect and collaborate, share each other's capabilities 02:42.070 --> 02:44.150 using a common protocol. 02:44.270 --> 02:46.350 So I'm really excited to show it to you. 02:46.390 --> 02:51.070 It makes a lot of sense to have that in week six, because we're going to draw on a lot of the things 02:51.070 --> 02:52.790 we've learned about in prior weeks. 02:52.790 --> 02:56.750 It's all going to come together, and that's very much the idea of this structure. 02:56.790 --> 03:02.590 Each week builds on the last as you build up your capabilities and you upskill all the way through to 03:02.630 --> 03:07.590 the end of week six, when you will have mastered engineering AI agents. 03:08.310 --> 03:16.510 And each week consists of five days, five different sets of lectures to build up your skill sets. 03:16.550 --> 03:19.110 During that week I give you Saturday and Sunday off. 03:19.310 --> 03:23.430 So just to show you what it looks like, I'm not going to talk to what we're going to do on every single 03:23.430 --> 03:23.870 one of the days. 03:23.870 --> 03:25.750 I want to give you that sense of, there's a lot going on. 03:25.750 --> 03:27.470 We're going to cover a lot of stuff. 03:27.590 --> 03:31.790 And this, this is color coordinated to match an earlier slide. 03:31.790 --> 03:33.670 That orange represents some theory. 03:33.670 --> 03:36.030 Building yellow is the project. 03:36.190 --> 03:39.630 And that first project we're going to do at the end of this week is going to be really cool, really 03:39.630 --> 03:45.390 fun, something that you'll be able to put into practice building your own agent that you'll be able 03:45.390 --> 03:51.910 to deploy into production, that we will do, that is able to answer career questions about yourself. 03:51.910 --> 03:56.030 It will be like your career alter ego that you can put on your website. 03:56.030 --> 03:57.630 Instead of having a resume. 03:57.670 --> 04:02.550 People can talk to you and talk about your experience and your challenges and so on. 04:02.590 --> 04:03.310 Isn't that cool? 04:03.310 --> 04:05.270 We're going to be doing that in the next few days. 04:05.590 --> 04:06.670 That's week one. 04:07.110 --> 04:07.910 Week two. 04:08.070 --> 04:10.630 We're going to be, as I say, OpenAI agents SDK. 04:10.910 --> 04:15.980 You'll see that the blue here, blue boxes represents a new framework that we are learning. 04:15.980 --> 04:22.420 And again, the yellow are the projects, including deep research, which will be a great app. 04:22.740 --> 04:24.580 Week three with crew. 04:24.780 --> 04:29.740 You can see we're going to learn about crew and then build a couple of different projects with crew. 04:30.100 --> 04:32.060 Week four is Landgraaf. 04:32.260 --> 04:40.980 Week five is Autogen and then week six is MCP and the capstone project, which is going to come at the 04:40.980 --> 04:43.500 end of this course, bringing everything together. 04:43.700 --> 04:47.300 So as I say, I'm not putting this up here to show you every single box. 04:47.300 --> 04:47.860 I'm doing it. 04:47.860 --> 04:50.700 So you see, we've got a lot of material to cover. 04:50.700 --> 04:55.540 There's a lot of areas in which we'll be growing skills, and by the end of it you'll have covered. 04:55.540 --> 05:00.540 You'll have ticked off everything you see here on this screen, the complete curriculum. 05:00.540 --> 05:05.020 And you will have true expertise in the world of AI agents. 05:05.540 --> 05:11.940 And I'm now just going to color in the the boxes which represent the delivery of major projects. 05:11.940 --> 05:17.740 And in particular, I would say that that the numbers five, six, seven and eight are some of the really 05:17.740 --> 05:19.580 exciting ones I have in store for you. 05:19.620 --> 05:24.460 I've just finished building them and I had so much fun doing it, and I can't wait to share it with 05:24.460 --> 05:24.700 you. 05:25.460 --> 05:30.420 And just to show you that, it's definitely worth hanging on in there for the next few weeks, let me 05:30.420 --> 05:33.220 tell you about those those last four projects. 05:33.540 --> 05:40.300 One of them is writing an agentic platform that's going to represent a whole engineering team, a team 05:40.300 --> 05:45.380 of different people, a front end developer, a back end developer, an engineering lead, and a tester 05:45.380 --> 05:50.660 that will collaborate together to write some software for us and we'll see them all in action. 05:50.900 --> 05:53.260 One of them will be, I call it sidekick. 05:53.300 --> 05:59.820 It's going to be an agentic platform that will be able to bring up a browser and interact in that browser 05:59.820 --> 06:00.940 together with you. 06:00.940 --> 06:04.460 So you will have your own sidekick that can do things with you. 06:04.500 --> 06:07.420 It's a bit like like a Marnus if you know Marnus. 06:07.460 --> 06:12.820 Of course, the Chinese agentic startup that's that's really caused again, a lot of excitement. 06:12.820 --> 06:18.260 But Marnus runs its platform in a sort of cloud box that you can't control. 06:18.260 --> 06:19.930 You can just watch it doing its thing. 06:19.970 --> 06:24.410 You'll have something a bit like that, but it will be on your box and you can work with it side by 06:24.410 --> 06:24.770 side. 06:24.810 --> 06:25.970 A true sidekick. 06:26.530 --> 06:30.930 Project seven I call creator, and this one isn't particularly commercial. 06:30.930 --> 06:36.690 It's just really interesting that this is one where we will build an agentic framework that is itself 06:36.730 --> 06:39.650 able to create new agents. 06:39.690 --> 06:44.690 It will actually give rise to agents and agentic platform that generates agents. 06:44.690 --> 06:48.650 And we're going to have those agents doing some commercial stuff just so there is a commercial angle 06:48.650 --> 06:49.170 to it. 06:49.610 --> 06:55.050 And project eight, the capstone project, it's something that I've wanted to build for so long, and 06:55.050 --> 06:56.850 I'm so happy to get to do it. 06:56.850 --> 07:04.370 This is a financial markets trading simulation where we will have multiple agents that can actually 07:04.370 --> 07:07.130 go off and make investment decisions. 07:07.170 --> 07:13.930 Searching the internet for financial news, looking up stock prices, real true real time stock prices 07:13.970 --> 07:22.090 and company information and read company annual reports, SEC filings and then make, buy and sell decisions 07:22.090 --> 07:23.570 trading on the market. 07:23.610 --> 07:29.250 Now that part is the simulated part, that we'll have some code that will just keep track of their accounts. 07:29.250 --> 07:33.610 And by the way, the code that we'll have keeping track of their accounts is code that will have been 07:33.610 --> 07:37.490 written by project five, by our engineering team, not by us. 07:37.530 --> 07:42.570 Just to add an extra dimension to this, some of the code would have been written by our Agentic engineering 07:42.610 --> 07:42.970 team. 07:43.370 --> 07:46.450 But what you see here in this screenshot is the results of this running. 07:46.450 --> 07:47.450 Just running yesterday. 07:47.450 --> 07:50.210 And it was actually a terrible day on the on the stock market. 07:50.210 --> 07:53.370 So all three agents lost money this day. 07:53.370 --> 07:58.930 But hopefully by the time that we get to week six, they will have recovered and we'll be able to actually 07:58.930 --> 08:02.050 show some profits that we're making in our virtual world. 08:02.050 --> 08:07.130 And of course, I should say now and I will say then that this should not be used for any real investment 08:07.130 --> 08:07.850 decisions. 08:07.850 --> 08:14.370 Of course, this is purely illustrative, but as an illustrative project, it's going to be a lot of 08:14.370 --> 08:14.810 fun. 08:15.450 --> 08:20.450 And while I hope that all of these projects will be entertaining, I will be trying to achieve two things 08:20.450 --> 08:21.090 from them. 08:21.090 --> 08:23.770 One of them is to have them be highly educational. 08:23.930 --> 08:28.880 Each project should allow you to upskill your capabilities and your expertise. 08:29.400 --> 08:31.880 The other is that I want them to be commercial. 08:31.880 --> 08:38.360 They should help you to have ideas of how you can apply this skill in business, whether it's a B2B 08:38.360 --> 08:40.080 or B2C kind of project. 08:40.200 --> 08:46.400 This should equip you to be able to put this into practice immediately the very next day on commercial 08:46.400 --> 08:47.120 projects. 08:48.000 --> 08:52.920 So now, just to introduce myself very quickly, just to show you that I am actually qualified to be 08:52.920 --> 08:54.120 talking to you about this stuff. 08:54.480 --> 08:55.800 So my name is Ed Dunner. 08:55.800 --> 09:01.160 I spent most of my career at JP Morgan, where I ran engineering and science teams of about 300 people. 09:01.200 --> 09:04.280 I started out in London, which is where I'm from originally. 09:04.320 --> 09:05.680 If you couldn't tell from the accent. 09:06.000 --> 09:14.320 And then I moved to Tokyo for a bit, and now I am based in New York and I am currently the CTO co-founder 09:14.320 --> 09:16.360 of an AI startup called Nebula. 09:16.400 --> 09:17.240 It's a nebula. 09:17.640 --> 09:18.680 You can go and check it out. 09:18.960 --> 09:23.560 And prior to that, I founded my own AI startup called untapped. 09:23.760 --> 09:27.720 And actually untapped was acquired in 2021. 09:27.760 --> 09:35.040 And this picture on the bottom left is a picture from the billboard in Times Square announcing our acquisition, 09:35.040 --> 09:37.720 which was a particularly magical day for me. 09:37.720 --> 09:38.200 So. 09:38.240 --> 09:41.040 So that was really, really an incredible moment. 09:41.360 --> 09:45.360 But I don't just show you this picture to show off about my moment in Times Square, but also because 09:45.360 --> 09:49.080 I happen to live about one block that away. 09:49.320 --> 09:55.680 If that guitar weren't weren't there in the hard Rock cafe, you'd be able to see my my apartment building. 09:55.680 --> 10:00.000 And I'm actually I'm facing the window right now that is, that is facing there. 10:00.120 --> 10:04.480 So in some way, if that guitar weren't there and this was some kind of a live feed, then, then then 10:04.480 --> 10:07.720 you would see me waving at you right now or something. 10:08.160 --> 10:14.560 Uh, another thing that I do is I love to be a speaker and instructor on all things LMN. 10:14.560 --> 10:22.680 And I have another Udemy course called LMN engineering, which is, uh, which has got 60,000 plus people 10:22.720 --> 10:25.200 on it and has been just an absolute joy. 10:25.320 --> 10:30.920 And I hope that some people that are on this now have actually taken that LMN engineering course, in 10:30.920 --> 10:36.190 which case, I must apologize to you because you heard almost exactly this same introduction when? 10:36.190 --> 10:37.350 When you took that course, too. 10:37.390 --> 10:37.990 So sorry. 10:38.030 --> 10:40.150 I hope you sped through it this second time around. 10:40.190 --> 10:42.230 Maybe it gets better on the second time, I don't know. 10:42.270 --> 10:47.430 Anyway, final thing to say, which I also said last time, is that here's a picture of me about to 10:47.470 --> 10:53.310 fly a plane, and you might think I'm showing you this because I am fabulously capable at flying planes. 10:53.310 --> 10:54.750 But no, I'm. 10:54.790 --> 10:55.310 AU contraire. 10:55.350 --> 11:04.230 I'm showing you this because my great skills in the field of LLM and AI is only surpassed by my complete 11:04.230 --> 11:08.510 inability to do anything requiring hand-eye coordination. 11:08.870 --> 11:14.750 So, in fact, if you find yourself in a plane and and you look in the cockpit and you see that it's 11:14.750 --> 11:20.270 me there in front of the stick, then you want to be looking for a parachute as quickly as you can. 11:21.150 --> 11:29.190 But but by contrast, if you find yourself taking a course about a gigantic AI and I am the course instructor, 11:29.350 --> 11:31.270 then you've come to the right place. 11:31.310 --> 11:33.790 This is completely in my wheelhouse. 11:34.350 --> 11:36.030 All right, let's get to it.