WEBVTT 00:01.220 --> 00:03.740 Well, I have to tell you that I'm a little bit sad. 00:03.740 --> 00:07.760 This is the beginning of the beginning of the end. 00:07.970 --> 00:10.940 It's the beginning of our last time together. 00:10.970 --> 00:11.780 The final. 00:11.810 --> 00:13.400 The end of the finale. 00:13.400 --> 00:17.330 The conclusion of our eight week voyage. 00:17.330 --> 00:21.860 To take you from the very beginning to LLM mastery. 00:21.860 --> 00:25.970 And what a fitting conclusion it should hopefully be. 00:26.000 --> 00:31.940 I realize I maybe oversold some of the excitement that was in store, but I do hope that you shared 00:31.940 --> 00:37.820 some of it with me, that you also enjoyed this project and that you're ready for the last stages. 00:37.820 --> 00:43.610 So what you can do now is so much to it that I couldn't possibly, possibly do it. 00:43.640 --> 00:46.010 The injustice of squeezing it into three bullets. 00:46.010 --> 00:48.590 So it now dominates most of this page. 00:48.590 --> 00:56.660 And obviously what's in store for us today is to get to the top of the mountain and complete the mastering 00:56.690 --> 01:05.990 of AI and engineering and LLM engineering, the missing piece for us, if we look at our Agentic AI 01:06.020 --> 01:13.040 solution, we covered each of these different pieces, and perhaps the one that we hadn't yet completely 01:13.040 --> 01:15.590 ticked the box of was having autonomy. 01:15.620 --> 01:16.520 We have got memory. 01:16.550 --> 01:21.560 We've in a sense, although there's not much point in having memory if it doesn't, if it only runs 01:21.560 --> 01:22.310 once. 01:22.490 --> 01:30.060 So yeah, we really need to pull all of this together and have it into a process that is running, and 01:30.060 --> 01:35.820 so that you'll be able to kick something off and then just go away and get on with your daily life and 01:35.820 --> 01:36.900 get texted. 01:36.900 --> 01:44.160 Push notified from time to time as deals come up on the internet that seem worthy of your attention. 01:44.460 --> 01:50.760 Uh, and so this one more time was the agentic workflows that we had put together. 01:50.760 --> 01:58.530 And of course, we built everything in red in one session when we we built out our, our prices, including 01:58.530 --> 02:05.310 the specialist pricing that calls out to modal and the frontier processor that uses our chroma lookup. 02:05.550 --> 02:10.260 Um, and then we built an ensemble that takes the weighted average of them all. 02:10.500 --> 02:16.590 Uh, and we then assembled it together with our scanner and our messaging and planning so that we'd 02:16.590 --> 02:18.060 have everything together. 02:18.060 --> 02:24.900 And then last time we used something that was fancily called an agent framework, but was nothing more 02:24.900 --> 02:27.750 than a Python script to glue everything together. 02:27.750 --> 02:32.880 And of course you can, as I say, get agent frameworks through through various off the shelf frameworks. 02:32.880 --> 02:38.670 But most of what they're doing is exactly what we did ourselves, uh, through simple Python. 02:38.760 --> 02:39.420 All right. 02:39.420 --> 02:45.660 So what remains, what remains is for us to go back with one last time to JupyterLab, where we will 02:45.660 --> 02:49.080 start playing with my old friend Gradio. 02:49.170 --> 02:51.090 Uh, we will be right there.