#6 Every Battery Is Different

0:00Christian Wederbrand: Welcome to the AI Gothenburg show.

0:07Viktor: Hosted by Dark Tranquillity and Red Bull.

0:11Christian Fleischer: Is In Flames allowed too?

0:13Viktor: Yes, yes, yes. They actually had an amazing show that week.

0:18Talare 1: Oh, oh, it was last week.

0:20Viktor: Yeah, you're the world winner.

0:21Christian Fleischer: Oh, I missed that. Me too.

0:23Viktor: It was actually a very good live concert with very good band, and also nice Gothenburg feeling. That's why that's what we love here. So, here in Gothenburg, physical hub, where you, Christian, have moved in. So we're gonna. Talk a little about more about you later on. Pretty soon, since we only do like 20 minutes. Yeah, and but you had something.

0:49Talare 1: No, Christian asked me who. Who the fuck are you? Oh yeah,

0:53Christian Wederbrand: And I was about to say it's better if Viktor introduces me.

0:56Viktor: I think because we had this amazing event in May, right? Hosted by Nvidia and Lovable, and lots of our current ambassadors and some of our. Members of the hub met actually at that event, and lots of fun things since then. And all of these ambassadors, which are pretty crazy, you're gonna meet a couple more of them here in the podcast moving forward as well. But it's so much energy, and we do so much cool stuff together. So it's also amazing to have you join the podcast and the team here. And yeah,

1:33Christian Fleischer: Thanks for having me.

1:34Christian Wederbrand: So I'm one of the crazy ones.

1:37Viktor: Yes, I know. Yeah. Yeah.

1:38Talare 1: Yeah. Understood. Sorry, sorry, sorry. But how interesting. Who am I?

1:45Christian Fleischer: I'm a co-founder of a company called Cognivity AI. We have been working in Cognivity for three years now, starting you know from the ground up, doing 50% consultancy, and then 50% trying to drive your idea. And our idea is that we build intelligent software for battery systems, and you know batteries are everywhere. You know you have them in your phone, in the car, In space, everywhere. So they are everywhere. But it's maybe still a little bit surprising that we don't really understand them still up to today. So we don't really understand what's actually happening inside when you, when you use the batteries, and how do you mean?

2:40Viktor: It's a chemical reaction. Correct. How hard can it be?

2:45Christian Fleischer: It's a chemical reaction which is very non-linear, and depending on how you use the battery, it will change its performance during the lifetime. So each battery is basically different. So when it comes from the manufacturing line, and how you're going to operate it. You will see different performance over the lifetime, and it is maybe different. Let's say it's take a use case of driving an EV. If you drive it here in Sweden, or if you drive it in Germany on a highway on the Autobahn, The battery will be used differently, and also the performance will change over this lifetime. So it's a huge market.

3:28Viktor: Lots of batteries out there.

3:30Christian Fleischer: Yeah. And as I said, you know, we some cases of how the battery is changing over the lifetime is not fully understood, and how we operate the battery is basically done with yeah standardized software, and this leads to the point that we will underutilize the battery over the lifetime. So we are. Combining data-driven solutions AI with battery physics, so that training, The learnings, adaptation to your personalized behavior is controlled by the by the physics as well, not only by AI.

4:18Christian Wederbrand: So in this, it is all batteries. Yes. Yeah, every battery.

4:22Christian Fleischer: Okay. Yeah, all lithium batteries. Okay.

4:26Viktor: Cool. And who is the customer?

4:28Christian Fleischer: So we have a long history within automotive industry, but at the same time, it's quite difficult to get into automotive. It's very you know regulated, and but there are other applications like battery energy storage systems. What the big batteries we use in the grid to stabilize grid to trade energy and also very interesting now drones. For example, drones which are used for medical delivery or surveillance applications. Space we're looking into space. Batteries are you know we want to bring compute power to space, so you need batteries to store during the orbit rotation. That you store energy when it's dark, so batteries could be also more deployed in.

5:28Viktor: In orbit. Okay, so this discussion could go in several directions. Yeah, yeah. Which one do you want to take?

5:35Talare 1: No, any one.

5:37Christian Fleischer: The thing is,

5:38Christian Wederbrand: When did you join and move into AI Gothenburg?

5:43Talare 1: Why? Why of course.

5:45Christian Wederbrand: What? It's important.

5:46Christian Fleischer: Yeah. So we have. I mean, we are small team.

5:48Viktor: What was that? Yeah. Why the fuck are you here?

5:52Talare 1: Alright, let's go.

5:53Christian Fleischer: We. We you know we are a small team. We were you know sitting in Johanneberg at Chalmers, and they were different companies, but not really technical companies. So, when you walk over the floor, you meet at the coffee. I mean, there was little synergies, little things to talk about technically and company and how you help each other in a way forward. And then. I don't know. Someone invited me to apply for Gothenburg. Maybe it was within our incubator. I'm not sure who was it, but I remember sending it way before Christmas last year.

6:34Viktor: Wow, one of the early ones.

6:35Christian Fleischer: Yeah, so it sounded very exciting. I read about your mission, your goals you have, who you would like to partner with. And it sounded natural to me to apply because it's also within our vision what we want to do. So we applied, and then we got the email. Yeah, thanks for applying. And then there was a long time a radio silence, and then yeah, we got an email saying, "Wow, we got so many interest." In this hub, so it takes a little bit longer, and then Michael, your colleague, he reached out and said, "Oh wow, Christian, we finally got to your application, and we really appreciate what you're doing. It's something very unique, and from the technical point of view, something we have not seen before, and therefore we would like to talk to you." Yeah, and then we met here. I don't know when it was. Was it April? I think April. Yeah, and then we had. Two three meetings here, and we liked it, and then yeah, before summer we moved.

7:53Talare 1: Ah, nice, amazing,

7:55Viktor: Yeah, yeah, very, very early once. Yeah, you need to sell the PowerPoint nowadays. You can sell a website or an entire company.

8:05Christian Wederbrand: The thing is, I like how you tell the story because we have a lot of applications for AI Gothenburg. That's interesting to hear the your side of the story in this case. So yeah, hopefully we will do even more job in the application site because we have a lot of them.

8:26Viktor: Yeah, and it's really noticed we have on our Builders sessions now on Fridays we had 20 people last Friday, 20 people Friday before that. Today we're gonna do a little speed dating format which we can, we have an idea for doing it on commercial with like pairing together our companies with. Customers, potential customers, or larger players who want to experience what we have here. Yeah, so let's do something about that.

8:53Christian Wederbrand: Yeah, and this morning we had a great breakfast session, right?

8:59Viktor: Yeah, we had CreativeMornings. They had their morning breakfast here. Yeah, and we also met Marc Hillander at the AI Gothenburg AI after work, great guy as well, and also a community builder. So let's try to do let's. Do more together. That's real mission. Also,

9:19Christian Wederbrand: One of one of our coming guests, I would say, for the podcast, of course, he will be here later on. Yeah, sure. Yeah, yeah, sure. Yes, he always goes has his own costume. I like that one, the blue one.

9:33Viktor: We can talk.

9:33Christian Wederbrand: No, no.

9:35Viktor: I'm I'm a little curious of your predictions of the future because you're a startup, you're early, we haven't seen this particular idea before. And you come, come away. How and when and where do the lines combine of the future mobility or battery usage? When do does that move into our homes? When does it get applied in different industries? Where do your intersection meet in the future,

10:06Christian Fleischer: So to speak? I think many people are getting now more aware of okay. We have a battery now commissioned, mainly from China. What do we do with it? How do we operate it? And I mean, talking to people who operate these big battery systems, they understand now they have no deep understanding of how the system works. And after they install it, they realize, oh wow, we can even optimize the operation. Because they still have a lot of trouble understanding the system, A lot of faults within the software, and they lose a lot of revenue during the years. And now they get, I think, a better understanding that there is a way to improve that, and they are more open to discuss these changes that we can that we are allowed to deploy our software on their systems. And I think they're getting more aware of that batteries. It's not like a phone, or like a battery you buy at Coop. You know, if you have batteries, you need to make sure that they are safe and still you can optimize because they will eventually degrade, if you use them or not. So and they were expensive, so the idea is then to optimize the usage and you know. Improve revenue, extend lifetime.

11:38Viktor: But isn't this very measurable as well? I mean, you have this huge facility up until now, measured this, and then you install Cognivity software, optimized, like you said, and then you have yes better metrics. Yes, from the start, more or less.

11:57Christian Fleischer: Yes, but the problem is also that the operators are not in control of the software. So that would require them to ask the producer in China, for example, to install the software. So when we talk about this field stationary application, it's we still need to communicate with the manufacturer.

12:21Christian Wederbrand: I have a question. In Sweden, since a couple of years now, we had this boom in solar power and also the storage with batteries and stuff. Yeah, yeah, correct. So and that's of course usually it's B2C, but are you focused more on the B2B side?

12:37Christian Fleischer: Yes. Yeah, that's only way. Yes.

12:39Christian Wederbrand: Okay. Okay. Because I'm sitting here like I need to optimize my own batteries.

12:44Talare 1: I have to.

12:45Christian Wederbrand: So I need to. So that's a huge market also, of course. But yeah. Okay. Understand.

12:51Christian Fleischer: In this case, we always need to go to the manufacturer. When we talk to the automotive. Manufacturers, it's easier because they produce also the software in most cases. Then they can just you know deploy it when they you know install their software. So yeah, that's it's always like tricky to get to the to the point that we allowed are allowed to deploy the software.

13:22Viktor: But if you can make a wish, what do you need to take the next step? If we have an anybody in the audience here that can help you, what would that be?

13:33Christian Fleischer: It would be yeah, an operator of a battery system who is willing to test our software and see the benefits of using our software. Cool. And this could be then, you know, helping others to get familiarized and get to know about what we're doing.

13:58Viktor: How do you get in contact with you?

14:00Christian Fleischer: Come to AI Gothenburg,

14:06Viktor: Or find Christian on LinkedIn. What's your surname?

14:09Christian Fleischer: Fleischer, Christian Fleischer. Cool.

14:12Christian Wederbrand: I have one more. Yeah, sure. He's a co-founder, right? Yeah. So how many are you in the company now? Our co-founding.

14:18Christian Fleischer: We are two. The other co-founder is more like it's my wife. She's more in the finance part, so not too much into engineering. She still finds it boring, but she likes the finance of it. I hope. And then we have, yeah, some data scientists and machine learning engineers and embedded system engineers working on the product development. Still a lot of research, but I think we have something very good because. Not only do we do this insights into the battery at the same time, for us it's also important data privacy. A big difference to solutions today is that you send your data to the cloud and then you have a centralized AI. We are doing that differently. We're moving the intelligence where the data is, and then it allows us to have this personalized software for each battery. As I discussed in the beginning, each battery will be different. But if you roll out software today, you have one software for all of the different systems, and then it's not perfect, not optimized.

15:33Viktor: But I but I guess it's also a security issue because if you have larger facilities with this type of software on, and I mean if you if you're hosting it in the cloud, and you get a security breach or leak, you need to patch it, right?

15:50Christian Fleischer: Yes, correct.

15:51Viktor: And.

15:52Christian Fleischer: It's also, I mean, it's your personal data. As you as a driver of your car, you might not want to share your personal information to the cloud. So, our solution is that we can still share information, but it's not personalized information. So, fair enough.

16:10Viktor: So both the security issues and the GDPR issues. Yeah. Europe compliant. Yes. Nice.

16:17Christian Fleischer: Yeah. That's idea. Yeah. Nice.

16:21Christian Wederbrand: Yeah. I don't have. I have a lot of more questions regarding batteries, of course. Yeah, I like it.

16:27Viktor: Full your home installation.

16:28Christian Wederbrand: Actually, the thing is, there's a new law. There's a new law in. I don't know if it's a EU new law regarding the energy for bigger companies now, because a lot of. I have some friends who works in this space where they build big sites. With very large storage facilities,

16:54Christian Fleischer: They need software. Yeah.

16:58Christian Wederbrand: Because that's on the B2B side for sure. Yeah. So maybe we can. Of course, I will hook you up with them, and then you can talk.

17:07Christian Fleischer: Oh no! For sure, for sure.

17:08Viktor: I think we have.

17:09Talare 1: That's a very good match.

17:11Viktor: I think we have a lead for you as well. I already emailed you that. [pip]. So now everybody knows that. Yes, someone in [pip] out there.

17:21Christian Wederbrand: Yeah, but I really do think you should look into the real estate part in Sweden, because that's I think it's going to be interesting.

17:33Christian Fleischer: Cool, thanks, guys.

17:34Viktor: Any end ending comments? Are you joining our Builders Community today?

17:39Christian Fleischer: Yes, I will. I would like to do the speed dating as I see. I mean, a lot of companies have moved in. I mean, I was out during the summer, but in the meantime, a lot of, companies moved in, so it would be nice to finally meet them and talk to them. Awesome! Let's do it! Let's do it! Let's go!

17:57Christian Wederbrand: Once again, welcome to AI Gothenburg.

17:59Christian Fleischer: Yeah, thank you.

18:00Talare 1: Thanks for having us. Good luck! Good luck! Me too. Me too. Okay, so that's fine.

18:07Christian Fleischer: Cheers! Cheers! All is clear.

18:13Talare 1: All is clear. Okay. Auf geht's! Auf geht's! Yeah.

18:16Christian Wederbrand: Nice run. So that's how we roll here in AI Gothenburg podcast.

18:20Christian Fleischer: I think this is how you call, you know, dip in cold water. Yeah, yeah, yeah. Next year will be different, maybe.

18:26Talare 1: I don't know. It's really good. Perfect.

18:29Christian Wederbrand: The thing is, I'm talking like this. It's AI Gothenburg Swedish. We call it. Oh, English. Yeah. It works. So, yeah. Builders Community in. Oh, we're on time. Hmm.

18:45Christian Fleischer: You have still twenty four minutes to publish it.

18:48Talare 1: Yeah, maybe it takes.

18:51Christian Fleischer: Put a filter, you know,

18:52Christian Wederbrand: Like an age filter.

18:55Viktor: But what's what's your workflow from here? So you have this in Riverside now.

19:01Christian Wederbrand: Yep, I'm going to make the magic audio. Magic is very important. Magic audio, and then you.

19:09Viktor: What was the name?

19:10Christian Wederbrand: It turns it into more of a studio sound.

19:15Christian Fleischer: Okay. Yeah. Can I sound like Bruce Willis? No.

19:17Christian Wederbrand: Ah. Not today, Peter. No, but that's one. And in the beginning, I fix the beginning and the end, adding the intro and outro, and also removing small parts of the pauses. That's it. And then it, then I go for export. And then I parallel to that I also have sent the transcription to in this case Claude actually, and then I get all the texts. Okay. Yeah, and then I also start at the same time when I download the file, I put it into my Video Factory to create the reels, all measure.

20:13Viktor: But isn't the background already in the studio? Yeah. Why do you need an extra transcription from Claude? Or you send the transcription to Claude? Yeah, yeah. For the bio episode.

20:27Christian Wederbrand: Okay, cool. And then it's then it's done pretty much, and then I upload it to.

20:31Viktor: Aren't there softwares to cut the pauses as well? Yeah. Or do you listen? You don't listen to it.

20:41Christian Wederbrand: No, no. You can you can take out magic audio,

20:44Viktor: Remove pauses, publish. Yeah, done.

20:52Christian Wederbrand: But that's how the workflow is. And then I also now put the file or the video into the Video Factory, and so it creates reels about what's the Video Factory? Ah, pretty much it's a it's a Remotion connected to Claude, and also yeah, that's that's pretty much it. What I have now.

21:15Viktor: Okay, but Remotion takes code and makes video.

21:18Christian Wederbrand: I mean, in this case, it helps. I do it for with Fable 5. You put in the video, and then it video file in Claude,

21:28Talare 1: Yeah. Really,

21:29Christian Wederbrand: And yeah, and then it creates exactly how I want it to be created. That the camera in this case we are like this, right? Yeah. Now I need to fix so it becomes a real because now it's like this and we are three in the same picture. So now I'm when I talk, The camera will be on my.

21:49Talare 1: And then you can zoom in, and yeah,

21:52Christian Wederbrand: And then the captions is there with our branding and also the AI Gothenburg first picture for the for the feed and such.

22:03Viktor: Is it a MCP connection connector to Remotion,

22:08Talare 1: Claude, or how does it?

22:09Christian Wederbrand: Yeah, yeah, I connected Remotion to Claude in this case. So that's pretty much what I've done. The thing is, I've also done the Video Factory for sure with much more things in Connie. That's my operation system. But the thing is, that was too much. So now I've slimmed it down. So now I work like this, much better.

22:32Viktor: Everybody wants Video Factory.

22:34Christian Wederbrand: Nemo knows everything now.

22:36Christian Fleischer: Yes. Yeah. So what is Nemo?

22:38Viktor: Yeah. My assistant.

22:40Christian Wederbrand: Connie is mine, and his name is Nemo.

22:44Christian Fleischer: Okay. What is the underlying?

22:46Christian Wederbrand: I created my harness is Hermes, Hermes. Yeah, so then I created a lot of things regarding how I connected everything. My stack is so the harness is Hermes. Still recording? No? Ah, but we have it here.

23:03Viktor: Yeah, but it's a extra.

23:05Talare 1: Otherwise, you need to be well.

23:09Christian Fleischer: Extended version.

23:11Talare 1: No matter either. Yeah.

23:14Christian Wederbrand: No, but regarding the we talking, this is after hours, would say. So this is after work, right?

23:21Talare 1: Yeah, we discuss.

23:22Viktor: No, but how we make the episode?

23:24Talare 1: Yeah, the scene. Yeah, still interesting.

23:28Christian Wederbrand: No, but the thing is, we actually talked about how we cut the episodes, and how I'm talking about the how I created a Video Factory on my operation system, Connie. Your is your operation system is pretty much Nemo,

23:42Christian Fleischer: Or he's just he's just Nemo.

23:43Christian Wederbrand: Yeah, yeah, he publishes everything. So now, in a couple of minutes, I would say we will have this episode up and running. But the thing is, I've created my Connie on harness in this case Hermes, and then I connected a lot of different software, of course, connectors to the whole. So I created a full stack, I would say. So, but the thing is, when everything is very easy, I would say it's very easy to build very broad, and that's what I've done. So now I'm a monster. Yeah, a bit a monster. It's really, really, and it's also on my local server. We talked about this in the in the episode regarding how you do with the cloud or whatever, right? In this case, local. Yeah. And it works, of course. But the thing is, I'm not that happy because it's a little bit of slow. Because I like this, right? So now I've created another one. Another one. Yeah, I really do need that actually. And regarding that, we need we really need to have talk about local servers in the pod later on. I would say because they're. The money for creating your own server at home now it's like I checked yesterday GPUs.

25:15Christian Fleischer: Yeah, I mean you can spend twenty thousand dollars on a GPU now.

25:18Viktor: But this is also an interesting discussion I had with a couple of other founders here. Is it going local or is it actually moving to the cloud? Because in the cloud you can scale infinitely, and if you have, if you can pick your token cost and your model and your task and slim that with a different kind of model, you can build like a you can build a monster as well, but which could actually perform. So you have two different conversion here, and it's going to be very interesting to see. Yeah, yeah, it's it's on the application.

25:54Christian Fleischer: I mean, a robot who washes your dishes doesn't need to be in the cloud. You need to make real time. Oh, my glass dropped. I need to pick it up. You cannot wait two minutes until you send something and you get a new training model. So this you want to have it, you know, In your on your device. But if you train one time thing like a big model, then you can do that in the cloud.

26:18Christian Wederbrand: I think you really need to have both ways actually here. It's going to be really important because also of course we are we are working with customers right. And depending on what area or space they are in, they need to have local AI because they don't want it up in the cloud. So and in some ways, you can of course go cloud on this customer, but in the bigger areas in Sweden, it's very like oh you need to do like this. So then it's really nice to have that. Approach to, of course, we fix this also. Yeah, I talked about it before. It's not like a big ass battleship that turns like this. It's or hangar or wherever. But the thing is, we are the small boats with the special forces.

27:08Viktor: Yeah, we are.

27:10Christian Wederbrand: We're standing in our team, and it goes like this: agile. Yeah, correct. So, that's what I think and I know a lot of the. Bigger customers, the bigger companies, they are pretty much on the big boat still, and we are going to go like this.

27:30Viktor: So we need to put some extra engines on the big boats.

27:35Christian Wederbrand: Yeah, for sure.

27:36Viktor: Yeah, I had a different what could you say like epiphany this yesterday morning where I wanted to. Okay, I need to do. I need to, I need to have a second opinion on my, on my Fable prompt because they still missed a shot, and I don't want to continue reading all the specs like I do. So I just copy paste into Chatty, and Chatty...

28:05Christian Wederbrand: Yeah, in this case, it's OpenAI.

28:07Talare 1: Ah, this is absolutely.

28:09Viktor: But it became Chatty because in my transcriptions from my Plaud device, everything when I said ChatGPT became Chatty. So okay, let's fucking name it Chatty. So I copy pasted from one chat box to the other, and after a while, first it was very human-like language, right? And after a couple of prompts back and forth, when Chatty and Nemo understood that they were talking to each other, and Viktor was off the loop, it was like more and more technical, deep.

28:45Christian Wederbrand: Gibberish stuff.

28:46Viktor: I've seen this episode on social media two years ago, and it happened to me yesterday. And finally, after like 30 or 40 minutes, it was like: "Yeah, Chatty: good work, Nemo." Nemo: "All right, Viktor, up to you, press the button." And I press the button, and it was like, oh my, okay. So I bet the system that I don't need to press the button anymore. And I will deploy it on Monday. Oh, nice!

29:16Christian Fleischer: That's cool. Yeah, you have like an auto loop or what between them? Yeah,

29:20Viktor: I think there are probably solutions out there. I've seen I've seen stuff about that as well. I have an idea, and I think I can build it.

29:32Christian Wederbrand: Yeah, but the thing is, yeah, regarding that, we would say with a second opinion in this case. I think a second opinion and maybe a third opinion also is really good to have now because you usually you work with one AI, to code right? Yeah. Go Codex, Claude or Grok in this case because there are the top three, let's say. Then.

29:55Talare 1: How I do it?

29:56Christian Wederbrand: I do Codex now because I really like how I can work on my phone because I'm always out and about, and also nowadays only goes by the tram and also the bus because then I can work from my phone. It's perfect for me, and I go for Codex, and then I have Claude coming in from the side here, and also Grok from the side there to see that it really, really. That is that it's legit, I would say.

30:26Viktor: That would be the cliffhanger on Grok.

30:28Christian Fleischer: Yeah, for sure, Grok, because yeah,

30:30Talare 1: You have your yeah, we have no,

30:31Viktor: No, no, no. Lots of people here is interested in what you can do with that.

30:37Christian Wederbrand: Yeah, yeah, for sure. So yeah, once again,

30:40Viktor: What's your name? What's your agent? I mean, what?

30:42Christian Fleischer: Oh, yeah, I'm I'm also in between Codex and Claude. I use Claude for planning and the implementation is Codex, and then I use Claude again for reviewing it.

30:57Christian Wederbrand: Okay, and the database.

31:01Christian Fleischer: Database? That's that sounds 80s.

31:03Viktor: What does that mean?

31:04Talare 1: Database. Oh.

31:07Viktor: I mean, no, he's from the seventies.

31:09Christian Wederbrand: Yeah. Okay, but I usually go for the thing is I never tried anything else, but I go for Supabase in this case. What is that? Supabase is the.

31:22Christian Fleischer: Yeah, where I do you mean? Do you mean like the your like visual studio code or something?

31:29Christian Wederbrand: No, no, in this in this case, I mean if you go for yeah, I started with Lovable.

31:35Viktor: I started with Lovable too.

31:36Christian Wederbrand: Okay, yeah, and they were connected to Supabase. That's why I knew about Supabase. Okay, so that's when I started my own projects outside of Lovable. I liked Supabase.

31:48Talare 1: So that's why we.

31:51Viktor: Like this community because, okay, we might be super nerdy in one area and have a blank spot in the other. All right, we can complete you, Christian. Christian, you can complete us. Awesome.