WEBVTT 00:00.000 --> 00:09.000 So welcome to the AI Plumers Devroom. 00:09.000 --> 00:11.000 We're doing it for the second time. 00:11.000 --> 00:15.000 Last year, it was called Low Railway Engineering in Hacken, 00:15.000 --> 00:18.000 but through like this year and a half, 00:18.000 --> 00:21.000 the new term appeared AI Plum in, 00:21.000 --> 00:24.000 just like Linux Plum in, that's what we do. 00:24.000 --> 00:27.000 So welcome to the Devroom. 00:27.000 --> 00:29.000 First of all, let's thank you to the sponsors. 00:29.000 --> 00:30.000 So we have two sponsors. 00:30.000 --> 00:32.000 I found an eigenface. 00:32.000 --> 00:36.000 There are lots of stickers for free to grab them. 00:36.000 --> 00:40.000 So there would be like other activities 00:40.000 --> 00:42.000 and some presents for speakers. 00:42.000 --> 00:45.000 So we'll talk about it a little later. 00:45.000 --> 00:48.000 The speakers dinner is today in the evening as well. 00:48.000 --> 00:50.000 And there is a fringe event in Monday. 00:50.000 --> 00:53.000 So you all invited, but please register. 00:53.000 --> 01:01.000 It's also in the process, but not in the university. 01:01.000 --> 01:06.000 And we just wanted to acknowledge that this event will not be possible 01:06.000 --> 01:08.000 first of all without the sponsors. 01:08.000 --> 01:11.000 Because again, there's a lot of things that we need 01:11.000 --> 01:14.000 aside from people helping organizing and showing up. 01:14.000 --> 01:18.000 But obviously, it probably wouldn't be possible without the people organizing. 01:18.000 --> 01:21.000 So you just see two of us co-organizers here, 01:21.000 --> 01:24.000 but there's actually a group of more like six people. 01:24.000 --> 01:27.000 And you will see them at various places. 01:27.000 --> 01:30.000 So thank you so much to our sponsors and our co-organizers. 01:30.000 --> 01:33.000 If you think round of applause, yes. 01:33.000 --> 01:42.000 It's kind of funny because we're running it for the second year. 01:42.000 --> 01:46.000 And it's really cool that every time this Devroom happens. 01:46.000 --> 01:49.000 And the event happens like about two weeks before the Devroom. 01:49.000 --> 01:51.000 And then everything is just about that event. 01:51.000 --> 01:54.000 Last year, it was all about DeepSeek, you know, 01:54.000 --> 01:58.000 how the model appeared and changed, you know, how we think about AI models. 01:58.000 --> 02:00.000 And you know, open models in particular. 02:00.000 --> 02:04.000 These year, of course, it's the cloud as in, you know, cloud's bought. 02:04.000 --> 02:07.000 And, you know, mold bought nowadays, you know, 02:07.000 --> 02:09.000 exactly, open claw. 02:09.000 --> 02:11.000 Yes, yes, some people pay attention. 02:11.000 --> 02:14.000 It's kind of cool that AI is so quickly developing. 02:15.000 --> 02:21.000 But we're still here to talk about things that much more at the planning level, 02:21.000 --> 02:23.000 much more fundamental level. 02:23.000 --> 02:27.000 And that's why you will not hear about a lot of things that are really cool, 02:27.000 --> 02:29.000 but it's just out of scope for this Devroom, right? 02:29.000 --> 02:32.000 You know, maybe there will be other Devrooms in the following years, 02:32.000 --> 02:34.000 but this one is just not one of them. 02:34.000 --> 02:35.000 So again, training is out of scope. 02:35.000 --> 02:36.000 Pythors is out of scope. 02:36.000 --> 02:39.000 You know, close source proprietary hardware is out of scope. 02:39.000 --> 02:41.000 Some things that are sort of in scope, 02:41.000 --> 02:44.000 and we might actually cover them on Monday at this event. 02:44.000 --> 02:50.000 MCP, agentic hacking, and obviously open claw. 02:50.000 --> 02:54.000 What you hear for though is modern inference frameworks. 02:54.000 --> 02:56.000 And by those, I mean things like Lama CPP, 02:56.000 --> 02:59.000 which is, you know, first few talks in the day, 02:59.000 --> 03:02.000 tiny grad, GML, you know, things like that. 03:02.000 --> 03:05.000 Unfortunately, the guy who is number one, you know, 03:05.000 --> 03:09.000 number second, you know, depending on how you count quantization expert in the world, 03:09.000 --> 03:10.000 could not come. 03:10.000 --> 03:12.000 So we had to replace his talk, and unfortunately, 03:12.000 --> 03:14.000 quantization will only be covered in the hallway track, 03:14.000 --> 03:16.000 but it's definitely in scope. 03:16.000 --> 03:18.000 But then inference management and deployment, 03:18.000 --> 03:19.000 you know, sometimes at scale, again, 03:19.000 --> 03:21.000 will be covered kind of in the midday, 03:21.000 --> 03:25.000 and open inference, and sort of inference focused hardware 03:25.000 --> 03:27.000 will be the last portion of the day. 03:27.000 --> 03:30.000 So don't miss that because there's a lot of cool talks there.