Getting AiRO with Ingmar Jungnickel
For most cyclists, wind tunnel testing is unrealistic. It is expensive, difficult to access, and often impractical outside of elite teams. Field testing methods like the Chung Method are proven, but they require time, experience, and careful execution to produce reliable results.
So what if there were something in between?
In this episode of Marginal Gains, Josh Poertner sits down with Ingmar Jungnickel, founder of AiRO, to explore how modern computational fluid dynamics has evolved into a practical tool for real-world cycling aerodynamics.
Ingmar is not new to this space. He is a mechanical engineer who has spent years working in cycling aerodynamics, logging extensive time in Specialized’s wind tunnel, developing on-bike aero hardware, and working across both cycling and Olympic speed skating. Long before AI became a buzzword, he was already applying physics-based modeling to human performance.
The discussion dives deep into how CFD has changed in the last five years. Simulations that once required massive in-house computing clusters and days of runtime can now be completed in minutes using cloud-based systems. AiRO uses parametric human models to create a digital twin of an athlete, allowing dozens of position tests to be run in a single session with repeatability that rivals traditional testing methods for many riders.
What surprises even Josh in this conversation is not just the speed of modern CFD, but how it reshapes the entire aerodynamic workflow. Instead of a handful of expensive tests, riders and fitters can explore position changes, posture, and equipment choices rapidly and iteratively before ever stepping into a wind tunnel.
This episode is not about replacing the wind tunnel outright. It is about making aerodynamic exploration accessible, scalable, and practical for athletes who would otherwise never have access to these tools.
In this episode, we cover:
-
Why aerodynamic drag, not weight, dominates cycling performance
-
The real limitations of wind tunnel and field testing for most riders
-
How CFD has become dramatically faster and more affordable
-
What a digital twin means for bike fitting and position optimization
-
Why rider repeatability is often the weakest link in tunnel testing
-
Where CFD still falls short, including apparel texture and skin suits
-
Why the future of performance looks more like position coaching than one-time fitting
Episode Transcript
This transcript was generated automatically by production software and has been lightly edited for spelling and obvious contextual errors. Minor conversational artifacts remain.
Episode Transcript
This transcript was generated automatically by Riverside.fm and has been lightly edited for spelling and obvious contextual errors. Minor conversational artifacts remain.
Hottie (00:06.07)
Okay, there I'm getting recording locally, cool.
Hottie (00:12.542)
Okay, excellent. And we're gonna do a quick countdown Ingmar with a clap so I can get a little sync spot when I do finalize the audio. I do use a little bit of this video for promotional purposes for an Instagram post and like that. So the video will be used to a certain extent, okay? All right, this is the Ingmar interview for Marginal Gains in five, four, three, two, one.
Ingmar Jungnickel (00:24.739)
Mm-hmm.
Josh (00:36.33)
Three, two, one.
Ingmar Jungnickel (00:37.262)
One
Hottie (00:42.576)
Excellent. And you want me to kick it off, Josh? And but I'll let the smart people really have the room, I think, on this one. I've got, by the way, there's a huge document here, Josh. I don't know if you saw it, but there's a massive document in here with Ingmar's background. And I think you have time to read it now, but it's all in there. So.
Josh (01:06.312)
Yeah, mean, I will... pulling it up.
Hottie (01:09.438)
And like I said in the email, I think the goal here is to get to a question 11 about as fast as we can so we can talk more about what you're currently doing. Your background is obviously very interesting, but let's as best we can jump to question 11 there. So feel free to like rip through things pretty quickly, write your background and so forth.
Ingmar Jungnickel (01:33.781)
Okay, will do.
Hottie (01:35.742)
And what did we just pick up? Some type of odd audio.
Ingmar Jungnickel (01:41.71)
Not here
Hottie (01:42.452)
Is that you Josh? Did something come on your? I'm just gonna.
Josh (01:49.108)
I don't hear anything. Here, let me go on mute and see if it.
Hottie (01:55.154)
yeah, was you. Something kicked on in your room. and it just went away. I think so. Yep, yeah, I Ingmar heard it too.
Ingmar Jungnickel (01:58.829)
hehe
Josh (02:02.11)
They go away.
Ingmar Jungnickel (02:03.757)
Mm-hmm.
Josh (02:05.822)
that's weird. Yeah, it doesn't sound any different in here to me. But, okay, hopefully it's...
Ingmar Jungnickel (02:06.734)
Mm-hmm.
Hottie (02:10.506)
Okay.
That's cool. All right, let's get going because we have some hard stop times here. Where's my thing? Here it is. Okay, cool.
Okay, well, welcome to the show, Ingmar Jungnickel. Now, Ingmar, I wanna make sure that I'm pronouncing your last name and your first name correctly. And the name of your company, which I hear you saying is AiRO, it's spelled A-I-R-O. So give us some pronunciations here first. I'm an old radio guy, I need these things.
Ingmar Jungnickel (02:41.558)
Yeah, perfect. Makes sense. I always say ordering at Starbucks quite difficult for me. I'm always still one cappuccino is typically what my name is ends up being pronounced at Starbucks. But yeah, first name is Ingmar. Last name is Jungnickel, Jungnickel in German. Jungnickel, I've lived in the US now for a while. I've heard it pronounced all kinds of ways. I respond to everything. And then company, yeah.
Josh (02:51.422)
Ha ha ha ha.
Hottie (02:58.297)
huh.
Josh (03:00.586)
Ha ha ha.
Ingmar Jungnickel (03:06.666)
With the name, everybody's doing this clever little AI putting AI into everything. We are on this train too, partially because of our product side of things and the company is named AiRO. AAiRO. Yeah.
Hottie (03:18.525)
Okay, excellent. But you are in the aerodynamics business, so that's an important thing to mention here. Of course, Josh Poertner, and we know how to pronounce your name, Josh, unless we've been getting it wrong all this time, is here with us too.
Ingmar Jungnickel (03:22.038)
Yes. Yes.
Josh (03:33.502)
It's so good to be here with both of you. We were just chatting offline and Ingmar was like, we've met before and so I'm sorry. But tell, well I guess we kick a lot of these off with how'd you end up in our little nichey niche of a sport? Yeah, and take us from there to you and I meeting previously.
Hottie (03:35.413)
Yeah.
Ingmar Jungnickel (03:40.888)
hehe
Ingmar Jungnickel (03:57.238)
Yeah, yeah, I'll give you the quick walkthrough. All in all, grew up in Germany, got into sports through one of my closest friends. The first sport I ever did was competitive kayak racing. And we were pretty decent locally, but turns out on a national level, not quite that good. Our local competition was not the strongest. We chose going to nationals, did everything wrong, sports science-wise. I finished that last. Yeah.
had a lot left to learn there was kind of over the sport, but through this, I really wanted to learn on what went wrong. What could I do better? I've always been a bit of a nerd. I've always been into physics, been into the sciences. So this was my first step is really getting into sports sciences and then also picking a different sport because I was done with kayak racing. And at that point I was using cycling as cross training. So I was running my bike around. didn't have a lot of money. So I had bought this used half complete bike on eBay.
and had to bike on my Northern Germany Dutch Cruiser bike from one shop to the other with the other bike on the back. Not knowing anything really about drivetrain parts, any of this is like, what's even missing? What do I need? And learned about all the used shops and having to do things, everything on the cheap. I got really interested in it and yeah, got to know all the local shops. And so the first part that really combined to me was this interest in sports science and the equipment side of things.
And I started reading on all these kinds of things, reading on slow Twitch a lot, reading on journalism articles a lot, everything I could find, and then getting more into the physics. Back then, I think analytic cycling was the page of the time with all these calculators playing with these calculators. And the thing that I realized at that point was that there's this big opportunity in cycling that somehow almost nobody was talking about. And that's that the number one resistance in cycling is aerodynamic drag.
Josh (05:39.978)
Mmm.
Ingmar Jungnickel (05:55.446)
and not weight and all the bike companies were talking about weight. And the two people that I knew back then that were quite vocal in this direction that I just fully agreed on were you and Mark Cody at specialized. So I started just hitting you up, going to Eurobike, talking to everybody I could, picked your brain. Actually another person, all of this, somebody I ended up racing and riding a lot with was Dave Kursl at Feld. So also Dave and I, yeah, we've been riding track and road together.
Josh (06:06.611)
Mmm.
Josh (06:19.831)
okay.
Yeah, yeah, who's also been on the show. So, yeah.
Hottie (06:24.711)
Mm-hmm. Yeah.
Ingmar Jungnickel (06:25.006)
Yeah, great guy. And so yeah, in university essentially built my degree to do sports aerodynamics. So I studied mechanical engineering in Germany, took all the low speed aerodynamics classes, all the CFD classes, both from aviation, automotive, basic fluid mechanic classes, and became the research assistant at the local university. And on the side, I was doing a project with a friend to try to do
power meter error testing outdoors. Of course, the Robert Chung paper I had found, Robert, I lived in Berkeley, or I lived in the Bay Area, he lives in Berkeley. Another great influence of mine got to meet, show him his stuff. That's how I learned programming actually was re-implementing his paper and doing a temperature drift correction on this because I was testing in the early mornings when the temperature changed a lot. That was my very first ever programming project.
Josh (07:09.77)
Go well.
Ingmar Jungnickel (07:20.526)
And yeah, so from there was working on this quite a lot. built our own hardware. had pitot tubes, we were measuring tire temperature with accelerometers. And that project was going on when Alpha Mantis came out. And we realized, we build almost the same thing. This seems like there could be a business here. And from there, we showed it to the German Cycling Federation. They were quite interested in it. But one of the things that then came up that actually foreshadows a bit what I'm working on.
Josh (07:32.563)
Okay.
Ingmar Jungnickel (07:49.248)
is that they felt this hardware solution was way too complicated. We had like 12 different sensors, there were cables everywhere. Even a government agency with many engineers felt like, this is too cumbersome to use. So we simplified it a lot, essentially to an almost only software solution and used it for track testing for the German national team. And they brought me on to work with them on error testing for German national team.
At that point, I was still in contact with Mark Cody, kept telling him about it. The German team ended up doing a good jump in performance during that time. And that offered me the job at, at specialized, which then cost me to work with us racing, all the specialized teams doing that. And then over to Morgan Hill doing RRNT at specialized.
Hottie (08:33.515)
Let me ask you this, Ingmar, before you move on. Regarding getting to know Mark and Josh, how encouraging were they at the time about your goals to be a bike aerodynamicist? Like, this is a very fine field, so were they like, yeah, come on in, the pool's deep, there's enough room here for everyone.
Ingmar Jungnickel (08:37.303)
Mm-hmm.
Ingmar Jungnickel (08:53.516)
Yeah, exactly. think what were the quotes? One feedback I got is like, there's total five of us in the world. Good luck was one of them. And I remember I went to Eurobike with a friend who was an aviation nut and really wanted to go into working on planes. And I think it was either you or Mark, Josh, who looked at the other guy and said, you'll be buying his beers for the rest of your life.
That friend is now the bike engineer at FES by the way, so he also went away from the aviation side. yeah, essentially I was pretty clearly told, don't do it, this isn't a job, there isn't an industry here. And back then there wasn't really a handful of people working on it.
Josh (09:27.97)
that's funny.
Josh (09:42.763)
Yeah, it really is amazing. mean, I think back the arc of my career, I I spent the first big chunk of it. mean, people just thought we were dumb, you know, like nobody. I mean, the bike brands, the teams, mean, you know, we spent almost three years in Europe talking to, you know, the bike sponsors, the team owners, the directors, and they're just like, nobody cares. You know, it doesn't work. It doesn't matter. I mean, and this is...
Ingmar Jungnickel (09:52.129)
Yeah.
Ingmar Jungnickel (10:04.813)
Mm-hmm.
Ingmar Jungnickel (10:10.803)
Mm-hmm.
Josh (10:11.946)
like the late 90s, early 2000s. mean, you know, like we had LeMond winning the tour in 89. I mean, you know, it wasn't, it wasn't like this was 1982. But it was just crazy how it just like people so broadly still didn't believe in it mid 2000s. And then when that switch flipped, was like, everybody was like, oh yeah, well, clearly, of course, know, duh, right?
Ingmar Jungnickel (10:16.139)
Mm-hmm.
Ingmar Jungnickel (10:20.494)
Mm-hmm.
Ingmar Jungnickel (10:38.19)
Mm-hmm.
Josh (10:39.594)
Andy Orting always said, yeah, 20 years later, we're an overnight success.
Ingmar Jungnickel (10:45.644)
Yeah, I was surprised how quickly the switch flipped, right? It went from nobody cared to everybody cared and nowhere in between, right?
Josh (10:49.609)
Yeah.
Josh (10:53.194)
Yeah, and it felt like, because I think we'd been grinding at it for just years and years and then yeah, you just kind of woke up one day and was like, shit, that really happened. Like in this new world. It's crazy.
Hottie (11:05.483)
Yeah, and what do you, for both of you, what do you think that pivot point was? Was there a moment in time where you went, oh, that's it, that's what flipped the switch?
Ingmar Jungnickel (11:08.302)
Mm-hmm.
Josh (11:17.854)
Ooh, I think it probably depends. mean, you have experience with like on the track side of things. I think certainly at the World Tour for us, it was maybe that second year with CSE where, you know, all of a sudden you had one team on aerobikes with aero wheels and nobody else was and they went from second to the last to like, you know, I think they finished the year second in the UCI points, but we're at one point.
Ingmar Jungnickel (11:24.878)
Mm-hmm.
Ingmar Jungnickel (11:38.382)
Mm-hmm.
Josh (11:47.147)
bumping the, you and you just kind of realize it's back to that marginal gains theory of like, well, if every rider on average is finishing eight places higher in every, you know, I mean, it wasn't like they were winning everything. They were just on the whole, like the math was mathing, you know, as we said, and then they were winning some races that, you know, maybe the otherwise wouldn't have or shouldn't have. I, yeah, I think.
Ingmar Jungnickel (12:03.566)
Mm-hmm.
Josh (12:15.978)
That was really notable. the other one I always say is the Perrubé win for the 303, for my career, that was the worst idea ever. It's never going to work. You guys are so stupid. And then literally the next year, it's like the French teams are like, ah, but of course, clearly, the carbon. And it was just like, oh, man, the whole business did a 180 in a season. And Arrow was probably, I think in Road, it probably happened over what?
Ingmar Jungnickel (12:33.795)
Hehe.
Josh (12:45.022)
two years and track, I think it happened much quicker because it was, it's just so much more immediate. You have some new tech show up and you're breaking a team pursuit record and then every other team, literally at like four minutes and 10 seconds is like, well, we got to do that.
Ingmar Jungnickel (12:48.27)
Mm-hmm.
Ingmar Jungnickel (13:04.334)
Trek was a great proving ground for me for exactly that reason. It's essentially a lab experiment. We did the error testing and we told the team you would go 3.6 seconds faster at the next World Cup. They ended up going 3.3 seconds faster at the next World Cup. And from that moment on, we had a pretty good trust relationship with each other. And on the roadside, it's just much harder to prove it that way. So many more variables go into it and it's a little bit more cluttered between all these other factors.
So I think it's harder to prove on track on the German Federation. They really got it at least on the endurance side on the sprint side, zero interest. I tried to get people to bend their forearms. I was told you cannot bend your forearms as a sprinter. Do you know how much power we put out? Cannot be done. I work with a US women's team sprint right now. Everybody's bending their forearms. It out 15 years later, what absolutely could not be done. Heavy pushback. Everybody is doing at this point.
Josh (13:32.394)
Yeah.
Josh (13:42.615)
interesting.
Josh (14:00.029)
Yeah, wow. When you look at the, I'm always amazed. mean, it's like these super narrow track bars. I mean, that was one of, know, first one of those I saw thinking like, God, you're never gonna convince anybody to ride that thing. It just looks dangerous. And then, I mean, you get one or two wins and then the next World Cup, okay, everybody's got them,
Hottie (14:00.331)
Hmm.
Ingmar Jungnickel (14:06.114)
Mm-hmm.
Ingmar Jungnickel (14:13.375)
Mm-hmm.
Ingmar Jungnickel (14:25.228)
It's what I really like about our line of work, it's very meritocratic in the sense, okay, people can believe in it or people cannot believe in it, but you'll just find somebody that does believe in it that you can work with and that person's gonna start beating all the other people and then they'll come along as well. I think that's the best strategy I found. It's hard to argue with winning.
Hottie (14:45.899)
You also have some very interesting time with speed skating. Tell us about that real quick.
Josh (14:46.014)
this for sure.
Ingmar Jungnickel (14:52.49)
Mm-hmm. Yeah. So that's our, that's what caused my newest adventure in all of this. So when I moved to the U S I had before worked with professional athletes, both on the Olympic level and to the France level site. And when I got into product development at specialized in Morgan Hill, I missed the interactions with athletes. And at that point specialized was the skin suit R and D partner for under armor for you speed skating. And I was the project engineer there.
So had the connection to speed skating. met Shane Domer, high-performance director there, and mentioned that I'd like to be still involved in an Olympic movement, and they have the Sports Science Commission, and he asked me if I wanted to chair that sports science group. And so from there, I started on the side with specialized. It was not competitive, right? So no conflict of interest. I chaired that sports science group, and we had a few projects go quite successful within it. And the most recent one is we built a computer model of the team pursuit.
We showed that there's a team pursuit strategy that is substantially more efficient than what was done instead of taking turns pushing essentially the entire way. And the US team has since then won every single World Cup, broke the world record and going back to it's hard to argue against winning every single team at the Olympics that meddled, did the new technique. And the Dutch team that had three individual world record holders in a team of three was the one team not to do it and finished fourth.
Josh (15:53.01)
well.
Josh (16:18.022)
Wow, that's impressive. that I love speed skating. It's such a beautiful sport. Yeah.
Ingmar Jungnickel (16:22.7)
Yeah. And the physics translates great. So the sliding friction coefficient is, you know, in the ballpark of what the rolling resistance is in cycling. And the drag coefficients are similar too. So a lot of the models I built for cycling, you can adapt pretty quickly. And there's a lot of technology transfer that we could do from cycling to speed skating. And now we're actually doing from speed skating to cycling.
Hottie (16:43.102)
Excepts.
Right, except skates have no power meter. So how do you factor for that? How do you figure out if the work being put out is being used as most efficiently as possible?
Ingmar Jungnickel (16:57.898)
Exactly, and that gets to the point of what I'm actually now working on and of this technology transfer from speed skating to cycling. So the German Federation and the Dutch Federation have power measuring skates, actually, but it's not something that is commercially available. We've looked into it and the US Federation being quite a bit smaller hasn't had the budget so far to work on this technology. But of course, that was our first question is can we get to power measuring skates? And with the sports science group, this is always one of our topics every year.
Hottie (17:09.449)
they do.
Ingmar Jungnickel (17:27.178)
It's are we ready or can we do it this year? With that being done, okay, what else can you do? Of course, one thing is to measure power, but then the other thing is to improve aerodynamics. And we've actually seen in cycling and individual pursuit that most of the gains we have seen over the last decade have not come from increased power outputs, but have come from reduction in drag. So we are pursuing that same strategy in speed skating.
And the other opportunities that we had is wind tunnel testing, which is hard in speed skating because you cannot statically hold those dynamic positions. So we tried to have them hold statically some positions they can hold, but they're unnatural for them, repeatability is poor. You can do it with mannequins, but mannequins don't really work for posture. And so we explored other techniques. And another project we went into was using a different approach called wake deficit.
So looking at the draft behind the rider. And in fact, if you look at the air behind the rider, you can measure the drag that way as well. This is the article that we recently saw from Bara in using PIV, Particle Image Velocimetry. And air bubbles essentially. we looked at, there's a Dutch paper called The Ring of Fire that influenced this approach. So we were looking at that.
But we're quickly told that in the US from a liability perspective, you're not going to have any Olympian ride through a high powered laser. And that's not going to happen.
Josh (18:49.852)
You
Hottie (18:54.933)
Hmm.
Ingmar Jungnickel (18:56.012)
So going from there, we explored other techniques and one project I had and one approach we were pursuing in around 2019 was an approach that is ultrasonic tomography. And essentially the idea is there's an ultrasonic wind speed sensor. So we measured the speed of the wind by measuring the speed of the sound going through the air, which travels at the speed of sound plus the speed of the air. So we can
back calculate the speed of the air. And if we bounce a sound in many different directions through the wake of a rider and we pick that up, we can use a bunch of math to three-dimensional re-create the wake behind the rider. Similarly, how a CT scan sending many X-rays through a skull, for example, can replicate that 3D model.
Hottie (19:45.544)
Somebody froze it.
Ingmar Jungnickel (19:51.938)
Then once you have that 3D model, you can replicate the wake. And this was my pre-pandemic project with two friends in the Bay Area that then unfortunately the pandemic killed, but it ran into that same problem as everything else that we'd seen is it's just too complex and too expensive. It's not even that the physics is the hard part, but it's that usability and the economics are the hard part.
Which then brought us to the last solution in all of this, and that is computational fluid dynamics and flow simulations. And what we had seen there and what we've used initially, and I know there's many people pursuing similar projects, is to combine this with 3D scanning. So you 3D scan an athlete, you then run CFD on it. The issue is that these scans are not very repeatable in their technique. And so they're labor intensive, you have to do cleanup.
You have to then rerun the simulations and repeatability isn't great. The one other option that we then found, which is our breakthrough is this parametric human model that we're now using. So we have a 3D model of a person. We can adjust around 4,000 things about this person. And it comes with an AI model that automatically tunes this parametric model to a photo and a few measurements. And so now we can create a 3D model of an athlete.
in around 10 minutes and do it fully automated. And then we can use these techniques from the special effects industry essentially out of gaming or movies to rig and animate this 3D model. So we can bend it into different positions and use motion capturing techniques or pose estimation techniques to recreate the bike pose. And this is what we used in speed skating now as the technique to do aero optimization is to optimize this.
digital twin we call it, this surrogate 3D model of the athlete to improve the posture.
Hottie (21:48.395)
One second Josh, whatever you did before do it again that mute thing or whatever you did something kicked on
Josh (22:02.356)
still hear it.
Hottie (22:03.261)
Yeah.
Ingmar Jungnickel (22:03.342)
Mm-hmm.
Hottie (22:07.211)
It's almost like you're...
Josh (22:07.466)
Yeah, might be the eighth there's like an it's probably the AC in the room adjacent because I do hear like a little bit of wind noise
Hottie (22:17.229)
And then it goes away. It's weird. Yeah.
Ingmar Jungnickel (22:17.422)
What's better?
Josh (22:21.086)
that's weird, because it doesn't sound any different to me.
Hottie (22:23.883)
It's almost like you got an automatic gain kicking on, some type of automatic gain control, but it's all right. It's fine. It went away. Hopefully it'll stay away till the end. And I can get it out too. It's no problem. So, okay. So we were at CFD. Go ahead, Josh, did you have a follow-up? Go ahead. You probably do it. You probably have a million follow-ups. Hit it. Okay.
Josh (22:31.347)
Okay.
Josh (22:39.942)
Okay. Yeah, yeah, no. So many follow ups. Okay, so, you know, this is fascinating. I have been out of CFD for about 12, 13 years now. So kind of looking in peripherally and still in touch with, you know, Matt Godo and some of the classic characters. But this is amazing to me and the more I read up on it. So you've
Ingmar Jungnickel (22:43.651)
Heh
Ingmar Jungnickel (22:55.054)
Mm-hmm.
Ingmar Jungnickel (23:02.957)
Mm-hmm.
Josh (23:09.488)
You've built this model that you can essentially turn in to the athlete by manipulating it. I think people who don't do this have no idea. like, you have to mesh these things. You're like wrapping this mesh around it and like there's like, you know, 20 million triangles. And if you're missing one triangle, the entire thing is like explodes and it doesn't work. you're I mean, I think I think the technology is getting better and better.
Ingmar Jungnickel (23:11.203)
Mm-hmm.
Ingmar Jungnickel (23:16.461)
Mm-hmm.
Ingmar Jungnickel (23:26.741)
Mm-hmm.
Ingmar Jungnickel (23:36.366)
Mm-hmm.
Josh (23:39.135)
But I mean, we would, in the old days, spend 10 hours meshing the thing. And then they'd be like, there's four triangles missing. You're screwed. so I guess, yeah, this just sounds amazing to me that you can do this. So I guess talk through some of the tech of having this manipulable model. should just consistently mesh every time, because you're just tweaking it. And then are you able to also use like, you know,
Ingmar Jungnickel (23:46.732)
Mm-hmm.
Josh (24:08.274)
rotating regions and movement of the model to do dynamic simulations, I'm thinking, yeah, talk us through how complex this has gotten.
Ingmar Jungnickel (24:19.214)
Perfect. Yeah. I think the, and I remember the Godot papers on safety, very influential for me. My thesis in college was the stability of side winds of wheels and crosswinds. so, yeah, read those papers a lot, worked on it. Yeah, exactly.
Josh (24:33.736)
Yeah, that was a huge, that was landmark, right? mean, that was, yeah, for so many of us were like, my God, it's, there's shedding happen. I all this stuff that we kind of, you know, whatever, you're like, it's, we can now see it. Okay.
Ingmar Jungnickel (24:39.585)
Exactly.
Ingmar Jungnickel (24:45.312)
Yeah. And we built on that in university. so we built, actually, we combined it with a vehicle dynamics model and we actually did the rider in the loop dynamics meeting a wind gust aerodynamics, turns into forces, forces into motion. Human controller stabilizes the bike. So on that side, you can get quite complex in all of these simulations. And I mentioned this too, because there's generally two philosophies in CFD. You can get higher fidelity and you can try to capture more more effects.
more and more accurately, or you can go broader and be able to do parameter studies and search your design space much more. And so you would run more mid-fidelity CFD, but much higher parameter studies around it. And my approach all the way back to when I let R &D development at specialized was working on the bench and working on those products was always much more focused on these parameter studies and exploring a space. So my
approach has really been okay. What do we want? We want to make athletes faster. What do we need to do? We need to find new positions that make people better and so go backwards from there. What do we need to do to do that? And it turns around, for example, in speed skating, athlete buy-in is so critical in all of this. So if you can do this in a session with an athlete where you can make it interactive for the athletes, that they can adjust the position and see the result in the same session.
That is really a breakthrough. So for us from the beginning, we said, we want to be able to turn the whole thing around within that same one hour session. So that was a critical part in all of this. On the technical side, what that meant is we had to fully automate every step. And this is where the parametric model came in. So both the pro and the con is this parametric model always has two legs, always has two arms. There's always five fingers on each hand.
There's always a head, the head's always in the position we would anatomically think it should belong. And of course, all these prior assumptions, knowing what a human generally looks like, allows us to fit these models much better with much less data compared to just using 3D scanning. So that's one element. And then the other thing is because it's already a mesh and we're just deforming this mesh through these parameters, we already have that initial mesh and this meshing step is much more trivial around it.
Ingmar Jungnickel (27:09.548)
So these two parts really meant that we can do that much more robustly and much faster and fully automated. And those were steps we needed in speed skating. And then later on, it also showed that those were the steps that we needed to make it viable for bike fitters, which was the issue we had faced 15 years prior, that all the techniques we found were just not economically viable from a time and effort perspective for bike fitters. So...
Josh (27:32.511)
Hmm.
Ingmar Jungnickel (27:36.608)
Yeah, speed and automation was a key piece for us in the ability to turn around these results.
Josh (27:43.903)
No, that's fantastic. mean, God, if you're just able to manipulate the mesh, because that's so much computing power and so much time and so much, you know, it's a lot of these things that the mesh is as much as running the simulation, right? And then you change one thing and you have to remesh the whole thing. so that I'm getting excited just hearing you talk about it. Question, you mentioned buy-in, which
Ingmar Jungnickel (27:51.401)
Mm-hmm. Mm-hmm.
Ingmar Jungnickel (27:59.49)
Yes.
Ingmar Jungnickel (28:06.222)
Mm-hmm.
Josh (28:10.398)
we've talked about on the show before, right? mean, you can have the best solution in the world and if, you know, you don't get, if the Spanish guy thinks it's gonna make his back hurt, he's probably not gonna ride it.
Ingmar Jungnickel (28:11.914)
Mm-hmm.
Ingmar Jungnickel (28:19.043)
Mm-hmm.
Josh (28:23.986)
Since you've worked both with what you're doing now and wind tunnel, do you find this is maybe an easier sell to the athlete because it's more visual and it's more interactive than the wind tunnel? My guess is going to be that it is because the wind tunnel you're like, yeah, I moved a little bit. You told me this number. They never really believe you, I feel like. This seems like it would be easier to sell. Has that been your experience?
Ingmar Jungnickel (28:52.362)
Yeah, I think they both have their pros and cons. So what I've seen with the wind tunnel is that athletes have to fly there. They have to go there. They know we're spending boatloads of money on them in this. And they're very present and they take that very serious just from this perspective. You can kind of coach it that way too, to say your team is dedicating a lot. You better make the most of it. So just this big structure, this, you know, the costs involved in all of this, it feels very space age, you know, aerospace side.
Josh (29:21.001)
Mm-hmm.
Ingmar Jungnickel (29:22.274)
That has its own pros and cons. While, you know, for me, for example, with speed skating, I'm there every Friday. And so we just do this more and more around it. So just being easy to use and being there all the time. That's ultimately how you build relationships and that's how you build buy-in. So a big part there, and I was actually told this too from one of the athletes is, Hey, we talk about technique all the time, but we go to wind tunnel once a year, and then some arrow person tells us to do something and then we forget about it again.
And just me being there every week is like, the arrow guys here, we got to make sure we're working on our posture. So even just that it's easier to use was an advantage. On the visual side, yes, we absolutely saw that. So initially we only output CDA and we had some level of visualization in it, but very rudimentary. And for us, what was the breakthrough is that we built this very nice particle visualization into our app. And you have this virtual smoke wand that you can move around and interactive and you can see the air.
And telling people, we ran a supercomputer and we simulated the airflow. It was all a little abstract, but now we just showed that visualization. like, this is what the app does. And that made it much more understandable for athletes to say what the app does and how it works. So yeah, buy in visualizing it. And my co-founder in the company, he comes from the gaming industry. This is his expertise is these visualizations, these animations. Turns out that's been actually quite helpful.
Josh (30:21.834)
Mmm.
Ingmar Jungnickel (30:51.616)
in making sure athletes go faster.
Hottie (30:54.943)
So does the model that you create, this parametric human model, does it look like a person? Like, if you did me, would it look like me on the screen? Would I recognize myself? OK. Yeah.
Ingmar Jungnickel (31:02.976)
Yes, yes, that's another critical part for us is that if Dathley doesn't recognize themselves, Byton goes way down.
Hottie (31:11.819)
Absolutely. So tell me what all I does it shows the athlete on the bike in the position that the app thinks be optimal? Am I getting all that correct?
Ingmar Jungnickel (31:26.358)
Yeah, I think to back up for a moment, just from the speed skating perspective, we had started in speed skating. We made this app just for the skating position. This is what your speed skating now uses going towards Milan Cortina here in a couple of weeks. And the way the agreement is, is I have the commercial rights for cycling and triathlon. So we spun off a business. This is AiRO. This is the company I'm currently working on and we're selling the technology to cycling teams and to bike fitters.
And yes, on this app, we are only focused on cycling. So we have a speed skating version and we have a skiing version. Those are only exclusive to Team USA. The cycling version and the triathlon version we sell. And the way it works right now, we really say it's a digital twin and a virtual tunnel. So what we can do is we can create your digital twin and we can create this experimentation environment where you can test many positions.
Right now the app itself does not give you any recommendations which position is fastest. It only tells you what the CDA of each position is. But of course, the name is iRODE, there's AI in our name. We have now a data set of over 4,000 positions tested. And of course we're running studies on all of this to say, can we predict which things are fastest? And then I have 15 years experience in testing positions in wind tunnel and track and outdoor testing.
And so we train our fitters with that knowledge as well.
Hottie (32:54.249)
How do we figure out which is fastest for the time being? Do you have to go out in the field and do that? Chung method, how do you do that?
Ingmar Jungnickel (33:00.086)
Yeah. So AiRO is another tool in the toolbox in that regard. Which tools we currently have available in cycling is wind tunnel testing, velodrome testing, field testing, CFD now in AiRO and looking at frontal area. And then copying the pros, which in reality, when you talk to fitters, wind tunnel testing, field testing, technically they can be done extremely accurate and they are very valid tools that we still recommend for pro teams to be part of your tool.
around it. The issue with them for most athletes is that they're too expensive and too cumbersome for people to use. So if you look what 90 plus percent of all athletes are using in the real world, it's either you copy what other people are doing or the most advanced thing that is like widespread is looking at frontal area and looking at people there. We looked at a data set of 1,800 positions in AiRO to say how much frontal area correlates with CDA and
That's the ANCDA, so you would expect a high correlation. But it turns out your CDA can differ by 25 % for the same frontal area. And of course, if your power meter is off by 25%, you throw your power meter away because it's a useless tool. So we ultimately concluded that frontal area as a tool is not the solution.
Hottie (34:00.268)
Mm-hmm.
Hottie (34:13.962)
Yeah.
Josh (34:23.306)
Yeah, that's funny you say that because it's it's a like an okay proxy and we, know, years ago, I remember doing that with you would take the picture and, you know, put it into Photoshop and make it black and white and we'd count the pixels and all that stuff. Yeah. And then you, you do kind of start to realize like, Oh yeah, it's we're, forgetting about the CD of the CDA and that those numbers can swing by, huge, huge margins. So that, um, so
Ingmar Jungnickel (34:28.631)
Mm-hmm.
Ingmar Jungnickel (34:35.977)
Mm-hmm.
Ingmar Jungnickel (34:45.111)
Yeah.
Josh (34:52.65)
I mean, it's fascinating. And I guess it kind of jogged my memory. I mean, think of, you know, in the old days, people also did what the pros did, but then the pros most of the time weren't actually wind tunnel testing what they were doing. It was the whole, was it like a two year period where everybody was like, AeroBar hands down because Ulrich did it and Ulrich won a time trial that way. then as we later studied, he had...
Ingmar Jungnickel (35:04.778)
Mm-hmm
Ingmar Jungnickel (35:11.338)
Mm-hmm. Mm-hmm.
Josh (35:17.994)
two humans worth of blood in that body. it probably wasn't the arms down position that got the job done there. But yeah, I how, I guess, what would you estimate teams using wind tunnel, mean certainly at the World Tour, probably 75 % of them? Like.
Ingmar Jungnickel (35:20.11)
Hehe.
Ingmar Jungnickel (35:23.85)
Mm-hmm.
Ingmar Jungnickel (35:35.278)
Yeah. Yeah. We're talking to a bunch of the teams and a few teams of them are using iRO at this point as well. And I think it depends on what your use case is. So if you're a World Tour team, what we recommend, or if you're a professional triathlete, is you can use tools like iRO as your pre-screening tool. So one of the things is we're 98 % cheaper than internal testing. We can test the position in seven minutes. And so we can, and we have for some athletes done 150 tests.
Josh (35:41.973)
awesome.
Ingmar Jungnickel (36:04.726)
in AiRO and we can use this before pro athletes go into the wind tunnel and pre-select which positions they should test. For world tour riders that have been to the wind tunnel before that are very repeatable in the wind tunnel, the wind tunnel is still the more accurate tool than CFD. But what's interesting in our case is when you have less experienced riders
Josh (36:12.798)
that's...
Ingmar Jungnickel (36:27.456)
repeatability actually goes quite a bit down because the main issue in wind tunnel testing is not the wind tunnel, it is the repeatability of the rider. So as you get more to the amateur athletes or athletes even that just haven't ever done an aero test, even pro athletes, young pros, top amateurs first time in the wind tunnel, repeatability is not as good as somebody that has been there in the past. And to give you a rough idea, that's the level of repeatability we can currently achieve with Aero.
Josh (36:29.066)
Yeah.
Ingmar Jungnickel (36:55.694)
So for us, for mid-level athletes, I should say mid-level athletes, those are people that want to win Ironman Hawaii and so on. It's not quite fair to call them mid-level athletes, but people that are not have the budget of a World Tour team behind them have not done multiple wind tunnel tests in the past. We can achieve similar accuracy. So it can actually be a replacement for wind tunnel for a certain subset of the population. And it can be a sp-
pre-screening tool for the most elite writers. And this is how we're currently using it.
Hottie (37:31.752)
Can AiRO help me make choices on equipment, wheels, helmets, skin suits, all these things we fret about on our online forums and here on Marginal Gains.
Ingmar Jungnickel (37:43.959)
Yes. So short answer in all of this helmets. Yes, absolutely. It's one of our most asked for features. We've scanned all the most common TT helmets. And actually what does quite well with our riders is to say, okay, we found the optimal position. Now let's tell you which helmet is fastest for this position. And in the past, I always joke, what was the only way to find the right helmet for you? You want to know which helmet to buy? Step one, you buy 15 helmets.
You then book a lot of tunnel time and then you got to get rid of your helmet. And our ability to test 15 helmets in three, four different head positions and do all of that in 30 minutes gives us the ability to really recommend which helmet is best for riders. So helmets, yes, absolutely. The second thing that we've done experimentally and that we get a quite interest in and will probably add to the app is bottle placement and triathlon. That's something triathletes are quite interested in.
Josh (38:38.634)
Mmm.
Ingmar Jungnickel (38:40.874)
And it turns out you can get quite large drag reductions if you place these bottles strategically. Things we currently don't support is skin suits. That's probably one of the largest limitations. This kind of modeling to model texture on athletes. There's academic literature now that shows it can be done. The main issue is it's between a hundred and a thousand times more expensive than the models we currently run.
So it's not necessary that we can't do it. It's just that we can't do it economically around it. And right now we build everything around this cost trade-off to be able to be fast and accurate enough for amateur athletes. One thing I should add, Josh, as you said, thinking back to CFD 10 years ago, I ran CFD at specialized five years ago, then took a time off from CFD and was quite surprised how much had changed. And this is the power of digital technologies.
CFD in the last five years has gotten 50 times faster and cheaper. And those are capability shifts, right? And we expect that in the next five years, it will get 50 times faster and cheaper again, while wind tunnel testing will still roughly cost the same as wind tunnel testing costs. So for example, we have a pedaling model already. We could do yaw. We are looking into texture. There's a lot of things we can technically do.
Josh (39:41.674)
Yeah.
Ingmar Jungnickel (40:01.548)
We just feel the cost and time trade off isn't there for amateur athletes right now. And for us, really the goal was to hit a price point that people can afford at their fit studio and the turnaround. we can run a CFD simulation in seven minutes and we can run many of them in parallel. And so our design goal actually was turn it around in five minutes. We're not quite there yet, but generally this was our key capability to say, okay, we want to be able to do this in a session.
Josh (40:16.82)
Wow.
Ingmar Jungnickel (40:30.294)
And then from here on out, we are investing all our resources in adding capabilities at that speed.
Josh (40:38.186)
That is so impressive. It took me through just, I mean, 15 years ago, I think we spent, I'm trying to remember, I had to get like mega approval to buy the computer through SRAM. But I mean, I think it was like a, I think it was 150, $200,000, couple hundred core, you know. And that was the beginning of like you, know, like cloud service computing. You know, this would have been like,
Ingmar Jungnickel (40:49.429)
Mm-hmm. Mm-hmm.
Ingmar Jungnickel (40:58.594)
Mm-hmm.
Ingmar Jungnickel (41:02.176)
Mm-hmm.
Ingmar Jungnickel (41:06.507)
Mm-hmm. Mm-hmm.
Josh (41:08.554)
gosh, 09, 2010. And so you were just, nobody had ever even heard of Amazon Web Service yet, because it was in its infancy. And so yeah, the capital investment was, hey, we need a $200,000 computer and then somebody to run it. And then it takes six hours to mesh and 20 hours to run the simulation. And so you're running, I I think.
Ingmar Jungnickel (41:10.602)
Mm-hmm. Mm-hmm.
Ingmar Jungnickel (41:17.966)
Mm-hmm.
Josh (41:35.199)
we thought we were hot shit because we could run like one simulation a day. And Matt Godo actually helped us build an automation that could, when it was done, it would just reload the next one and then mesh it. so you didn't have to like wake up, you know, because in the beginning it was like, I'm sure you've been there, like it's like your student days where you're like setting the alarm, like, it's going to finish at like 2 15 in the morning. So I'm setting this alarm because I only have access to it for so many days. So I guess.
Ingmar Jungnickel (41:39.746)
Mm-hmm.
Ingmar Jungnickel (41:45.321)
Mm-hmm.
Hottie (41:55.851)
you
Josh (42:02.94)
Yeah, and now you're talking about, it's seven minutes. It's in the cloud. We can run them simultaneous. What's the back end of that, technically? Because that sounds amazing.
Ingmar Jungnickel (42:06.976)
Yeah. Yes. So you mentioned AWS. That's exactly what we run for high performance compute. We use the largest computers that they have on there. So that was one of the first solutions is let's throw as much compute at that as possible. And to give you an idea of the scale, every bike fitter has in, if they run one simulation between two and three times the compute power I had at specialized when I designed the bench.
So when you press that simulate button in the app, you unleash two to three times of that of a large bike company that bought a large cluster to do this. And every bike fitter can launch 10 to 20 of those simultaneously. So just in raw compute power for those that are into computers, it's 192 core per simulation terabyte of RAM. Yeah, and the fastest machines we could get.
The second part in all of this is we tuned everything in that simulation for performance and turnaround time because that is what we're optimizing for. So one example we're looking for is how we're initializing the airflow within it. So the initial guess for the simulation. And because we only simulate athletes, we can actually turn to AI tools and learn tools and strategies around it because, you know, we're running very similar simulations every time and that can help us largely accelerate these.
Another decision that we made is in our customer facing simulations, we do not have a bike in there. We only test the flow around the rider. And that might be surprising how we made that decision, but the important thing is CFD, the speed is all tied to the mesh count. How many cells do you have in your simulation? And what drives the mesh count is very heavily the surface area of your model and then the tightest curvatures that you have on your
And the bike itself has many finer details than the rider. And it means adding a bike doubles to triples the cell count that you need in the simulation, which drastically slows down the speed of it. We have these simulation capabilities, of course, to do it with a bike as well. Just for a rider, we did the tests if we would recommend a different position with the presence of the bike, if the interaction effects on a classic aerodynamic shape, not like a Hope Lotus or a Look.
Ingmar Jungnickel (44:33.358)
Or maybe the new Factor bike, but on a classic aero bike, does the presence of the bike mean we would recommend a different handlebar? And the answer on this is no. So we are able to drastically speed up our simulation by not simulating the bike. Of course, that also means we can do recommendations on bikes and wheels. But at the same time, we see that the independent tests that you can find for wheels, this is not the part that's very personal dependent.
So the things that are very person dependent is the fit, the skin suit, and the helmet, and we can do two out of three right now, and we're looking into the third.
Hottie (45:10.803)
and socks, Josh sells a lot of Aero socks, you do my socks?
Ingmar Jungnickel (45:13.718)
Yeah, we can't do socks yet. So that falls into a peril, which is called transition modeling and texture modeling. This is one of the things we're actively looking into. And as I said, technically about to be able to do it, but to do it economically is going to be the main hurdle here.
Hottie (45:37.503)
Well, Josh, I'll have to wait for your socks to get the the AiRO approval.
Ingmar Jungnickel (45:41.589)
I think Josh, you muted.
Josh (45:44.725)
Sorry, I'm mute. The texture modeling alone, I mean, is challenging enough, but then you get into, I mean, anybody who's been in the wind tunnel with one of these suits, know, different writers wear the same suit and you get totally different results because of fit, because of wrinkle placement or presence or the seam falls here on this person, but somewhere slightly different on another person and you get a totally different result. And so, yeah, I could really imagine.
Ingmar Jungnickel (45:50.765)
Mm-hmm.
Ingmar Jungnickel (45:57.358)
Mm-hmm.
Ingmar Jungnickel (46:05.293)
Mm-hmm.
Ingmar Jungnickel (46:10.061)
Mm-hmm.
Josh (46:13.922)
I think for the – because we get questions like this I know from our athletes all the time. said there are some things that you're probably better off to not simulate broadly because they could lead you down pathways that are just wrong, right? Like that – because they'll come to us and say, I saw this Alex Dowsett video and this suit was amazing. Like, well, yeah, on him. But that doesn't mean it works on you, right?
Ingmar Jungnickel (46:29.518)
Mm-hmm.
Ingmar Jungnickel (46:37.322)
Mm-hmm. Yes. Yeah.
Josh (46:41.962)
So yeah, that's the details, right, where the money gets spent. And I'm sure it's exponentially. So 192 cores. OK, that's awesome.
Ingmar Jungnickel (46:51.51)
Mm-hmm. Yeah. And I think for us, a big part here is we want to help people go faster, right? This is always the first thing we say is what is the biggest impact we can have in people going faster. And this recognizes that we don't have certain capabilities and we have to make certain decisions around it. And for example, no, we cannot simulate skin suits. That's a limitation on it, but we can simulate posture pretty accurately.
Josh (46:54.77)
and a terabyte of RAM.
Ingmar Jungnickel (47:18.848)
and cost-effectively, which also means we can explore more positions. And so we try to make these design decisions to say not what is the highest fidelity simulation or what is the simulation, what could we all be doing, but really saying what are the decisions that help us make our riders or athletes as fast as possible. And in this case, turns out as we're on the Marginal Gains podcast, I think that's a part I should bring up as well.
One of my jobs with the US Olympic Committee right now is consulting on innovation strategy. And the strategy we're working on, we call the Pareto gains concept. And it's kind of counter-positioning to the marginal gains concept and saying we actually want to do very few changes, but we want to do the most impactful changes. And if we want to do the most impactful changes,
there's much more effort in exploration because we know they're not in the places where we've looked before. If we keep doing the same things, want, we all get the same results. And there is a strategy in aerodynamics to say, okay, we can fine tune. We can, you know, keep doing the same things we've been doing and better and better, or we can be much more exploratory in it. And so the strategy I've been pursuing in the strategy for reasons we can go into it that I'm advising with my customers.
is to do a much higher focus on exploration. And from this, we want to build tools that enable this exploration. And one of the nice things in working on a digital twin is you can test positions that an athlete can't hold yet. You can test positions that may have an injury risk with it. You can test positions that might distract an athlete from a current situation. And you can actually run these experiments away from an athlete first.
Josh (48:54.973)
Mm-hmm.
Ingmar Jungnickel (49:08.814)
then work with strength and conditioning, then work with their coaches and say, okay, can we come up with a plan in like three year time horizon, for example, if we're thinking towards LA 28 and come up with positions. And I think that's an important part where I'm coming from to say, okay, I want to build experimental tools and I want to build exploratory tools that allow you to think more out of the box and try things without.
Josh (49:16.638)
Mmm.
Ingmar Jungnickel (49:37.173)
massive costs per experiment.
Hottie (49:41.429)
Well, if you want to start the Pareto Gains podcast, let us know. mean, maybe we could start a side channel somewhere. Doesn't roll off the tongue quite as well, but I like the idea.
Ingmar Jungnickel (49:44.27)
Okay, sounds good.
Josh (49:52.327)
Yeah, yeah, it reminds me a little bit of, know, it's one of the only, I think one of the only athletes we've ever interviewed who kind of got into this was Sophia, that she talked about, you know, having like two and three year plans for certain things, like a position change, right? And then like planning to build into that. And I think, yeah, what you've just hit on, know, people don't think it's, sometimes you have these like regions of disconnected performance where,
Ingmar Jungnickel (50:09.911)
Mm-hmm.
Josh (50:21.106)
you spend all your time pushing the boundaries of this one little region of performance.
You know, for your average listener, probably the best example there would be like the firecrest technology that, you know, we all rims look the same and we tried thousands and thousands and thousands of them and then all of a sudden you had this one kind of backwards bulbous really stupid looking one that was amazing. You're like, man, we've never done any research over there. And then it turned out that there was this whole area of interest.
Ingmar Jungnickel (50:43.245)
Mm-hmm.
Ingmar Jungnickel (50:46.914)
Yes.
Josh (50:51.626)
had you just gone in a nice linear stepwise function, know, 2 % wider, 2 % bigger, 3 % here, you never would have gotten there. And so that's an amazing one, yeah, being able to test a position that the rider can't hold or maybe the bike isn't there yet. we've certainly seen it these last four or five years. I mean, these 140-millimeter stems, narrow, I mean, this long, narrow
Ingmar Jungnickel (51:16.792)
Mm-hmm.
Josh (51:19.738)
thing that happened that's been huge. you may have found that before anyone else if you had your model that you could kind of pull and stretch and go, well, nobody makes 150-millimeter stem, but if they did, what would happen? And then you can make that stem. Yeah, because the reality is nobody made that stem because nobody in their right mind ever would have written that stem until you had the data to know that.
Ingmar Jungnickel (51:34.058)
Yeah, and then you can make that stuff.
Ingmar Jungnickel (51:47.104)
Exactly. And I think that's the value of digital things, is you can do things that you can't yet do in the real world. At the same time, that's the downside of it too, is that in the wrong hands, recognizing that only because the CDA is low doesn't mean that that's what you should be doing. It's something that ultimately shows us to decide, okay, we're working with professional bike fitters and experts in the space because it does require some training around it.
Josh (51:48.446)
that works really well. That's amazing.
Ingmar Jungnickel (52:14.68)
But at the same time, yeah, one thing is you can enable much longer time horizons. Where I see the future going with all of this and what I talk to the fitters we work with is that if you want to get the maximum performance, I don't think of it really as much as bike fitting, but position coaching. And the comparison is no swimmer would ever expect that they go to like a swim fitter. That swim fitter one time explains to them what perfect swimming form looks like.
And then afterwards they just go and do it. And cycling is quite technical in that regard as well. There's a lot you can do in neck shoulder range of motion. There's a lot you can do in multi-year time horizons that get you much better performance results, but it's an ongoing commitment and you have to train it. For the people that want to get the best results, that's the way to do it. I also recognize there are some people that just want the best position in that fit session and that's also okay.
Josh (53:00.938)
Mm.
Ingmar Jungnickel (53:12.706)
But yeah, if you want the fastest, you gotta work on your position.
Hottie (53:16.587)
Hmm.
Josh (53:17.598)
Yeah, and I'd say beyond the digital world, I cannot tell you how many wind tunnel sessions I've been to where they end up with a rider in some position that, you know, it's usually me in the room going like, you know, he's never going to ride. Like that position is never leaving this room because they take it home and they ride it one time. it's like because there's no plan. You know, it's like, we moved you six centimeters and this way and that way.
Ingmar Jungnickel (53:30.602)
Mm. Mm-hmm.
Ingmar Jungnickel (53:41.771)
Mm-hmm.
Mm-hmm.
Josh (53:47.155)
you have zero rider buy-in or capability. I will give a shout out San Diego Botero, probably the most amazing athlete I've ever worked with from like a flexibility perspective that you could just put that guy anywhere. And Ekimov was pretty close, but you could put that guy anywhere and he could make power. I mean, we used him to do all sorts of crazy, kind of like this, like, what if you had a...
Ingmar Jungnickel (53:50.04)
Mm-hmm.
Josh (54:13.736)
you know, negative 30-degree long stem. was like back in the Phonak days and I mean we – it was amazing. Yeah, I can do this. I'm fine. And there's other writers. You move them a centimeter and they're like, you're killing me, So yeah, I think it's – we often forget that difference in the rider of like, hey, this might take you six months to achieve this faster position. you know, your swim –
Ingmar Jungnickel (54:22.784)
Mm-hmm.
Ingmar Jungnickel (54:27.342)
Mm-hmm.
Josh (54:41.962)
analogy is a great one. Just because you see it once or say it doesn't mean it's achievable.
Ingmar Jungnickel (54:50.166)
And in the professional world, that's one of the challenges that I've seen with wind tunnel testing is let's say you do test a more out there position and it is substantially faster. What has happened to me multiple times then is the athlete wants to do it from then on out, right then and there, right? And they switch into this position.
Josh (55:04.435)
Mm. Yep.
Ingmar Jungnickel (55:08.192)
And then I call 10 days later and they said, yeah, my neck hurt so much. couldn't sleep for two days. I'm back on the old position. And it's like, well, yeah, you tried to do a massive change from one day to the next and then completely overdid your training. so figuring out what is the fastest position and how do you execute that fastest position on race day are two completely separate questions. And ideally you first figure out what is the thing to do and then you make a plan.
Sometimes it's great if you can involve the athlete in a plan, if the athlete understands that it has to be taken step by step. Sometimes it's nice if you can do the exploration without the athlete there and then come up with a plan how to approach it with the athlete. And so there's pros and cons to every strategy.
Hottie (55:49.388)
Well, Ingmar, I imagine you're collecting a lot of data. Data seems like the new gold by today's standard, right? It's worth a lot and it can tell us something about where we're going or our future. California is perfect example of that, Valuable gold and it really laid the landscape for this state. Data seems like the new gold. So tell us, is there anything about the data you're collecting that can give us a little insight as to where we might be headed aerodynamically? Is there a new frontier that's starting to emerge?
Ingmar Jungnickel (55:55.214)
Mm-hmm.
Ingmar Jungnickel (56:01.006)
Mm-hmm.
Ingmar Jungnickel (56:19.082)
Yeah, absolutely. think the first part in this is these things that we talked about capabilities we currently don't yet have, for example, apparel or some of these other parts. One big thing we are currently working on is to use these physics informed AI models is what they're called to train a replacement model to replace CFD. And what our roadmap there looks like is that it's actually not that realistic short time that it replaces CFD.
entirely, but it can provide an initial guess to that CFD simulation. And then the CFD simulation will, what's called converge, it will arrive at its results substantially quicker. So this is our current part is to use our own data to accelerate our own simulations. And going back to, it's between a hundred and a thousand times more expensive on apparel. Going back to we've gotten 50 times faster in the last five years.
If you put these technologies together and this is just a mind blowing speeds of digital things, this doesn't mean a hundred to a thousand times is undoable. It actually means, that's next on a roadmap. So that's what we're working on currently. So this is foreign internal data and it will enable us to run CFD simulations, error simulations, just orders of magnitude faster and cheaper than somebody that doesn't have this data set.
The other project we're currently working on is just running models to understand posture. This is called like interpretability of results. So we have, I think, 28 variables in our app. So we consider like hip angle a variable, lower back curvature a variable, upper back curvature a variable, know, elbow with a variable, wrist with. So there's 28 things you can combine.
And the amount of combinations you can put people in positions is staggering. Like it just cannot be done to run all these combinations on it. And we know from wind tunnel testing and from experience that a lot of these are quite complicated interactions on it. It's like, okay, if your hip width is like this and your hip angle is this, then your bar width might look like this kind of situation. And so there are some of these rules that are pretty easy and pretty universal. And some of these rules that are a little bit more complex.
Ingmar Jungnickel (58:38.326)
And you can really unearth these rules if you have a larger data set. So now that's an ongoing project as well, working with a group of interns right now to say, okay, what does the data show us for fit recommendations that work? Like, can we come up with more simple rules of thumbs, for example, for bar width and say, okay, what are the variables that drive bar width? And in our case, it's very nice in a wind tunnel test,
It's even hard to describe these variables like hip width and so on. And you'd have to measure them all when the athlete is there. And then maybe a week later you come up with some kind of way, okay, but we didn't consider this variable, but you didn't measure it at the athlete test. In our case, because the whole thing is digital, we have every variable that perfectly described that situation. And so we can find these correlations that we then ultimately can use to recommend to our riders and our athletes.
Josh (59:08.98)
Hmm.
Ingmar Jungnickel (59:34.712)
how to arrive at faster positions.
Josh (59:40.425)
You got my brain spinning. mean that, yes. Because there's all the things that you see in the tunnel. You're like, well that writer has really big hands and so that hand position worked for him. So is that a big hand, you know, for an example or yeah, the wide hip thing. Like, did that work because of that? And you never really know. And so to have the data, the gold, is hotty.
Ingmar Jungnickel (59:42.168)
Hmm. Hmm.
Ingmar Jungnickel (59:53.63)
Mm-hmm. Mm-hmm.
Ingmar Jungnickel (01:00:02.026)
Mm-hmm.
Ingmar Jungnickel (01:00:07.66)
Mm-hmm.
Josh (01:00:10.054)
is calling it. Yeah, you have all this gold, you're like, well, let's look at the last 200 large hand writers and what would that be? Yeah, that's…
Ingmar Jungnickel (01:00:16.478)
Mm-hmm. Mm-hmm.
Yep, and even more so, what our parametric model enabled to us is that we can do what's called synthetic humans. So we can actually just vary hand size, for example, in your example. We can say, we've come up with a theory that hand size is really important in combination with this helmet, let's say. Yeah, we can create 200 humans that are only differing in hand size, and we can run them through AiRO, and I could give you the results tomorrow on all of this.
Josh (01:00:28.682)
Hmm.
Josh (01:00:46.3)
Wow.
Ingmar Jungnickel (01:00:47.086)
We're the client, we ran 1,600 simulations, 1,600 athletes, and to do these kinds of correlation studies, and we did that in two days. So this ability to run parameter studies and then generate these kinds of insights, that I think is the big opportunity to generate them and to really take the next step forward on it.
Josh (01:01:06.218)
Hmm.
Ingmar Jungnickel (01:01:10.154)
And yeah, on the scale side, we are, over 4,000 simulations now. I tried to estimate how many wind tunnel tests I've done in the last decade combined. And I think I've done around 4,000 tests on the order of magnitude, maybe a little bit less, maybe two to 3000. So I've already in the last six months run more simulations that I've done wind tunnel testing. We're expecting by end of 26 to have run a hundred thousand post simulations. And we're expecting that.
You know, by 26, 27, we will have done more arrow tests than the entire arrow community has done prior in history combined. So the ability to do these kinds of tests is what has me very excited. Yeah.
Josh (01:01:53.951)
the scale. mean, yeah, because you're right. It's – that you can do things in parallel and that you can do them so quickly is just – OK. You got my head spinning. All right. Follow-on idea if anyone wants to pick this up, but the cosmetic and orthopedic surgery to become the synthetic human that is faster than you. Just think about it. You know, it could be a whole follow-on business.
Ingmar Jungnickel (01:02:00.375)
Mm-hmm.
Ingmar Jungnickel (01:02:04.322)
Yeah.
Ingmar Jungnickel (01:02:20.3)
Yep. Well...
We had one team actually ask us about how much does biceps diameter matter in this? So we did run studies to say, okay, upper arm diameter with a track team, the question was, okay, we do strength and conditioning and the gym is discerning our performance. Should we actually do this? And so, yeah, we're not doing any cosmetic surgery kind of stuff, but in terms of, musculature, stomach in, stomach out, relaxing stomach muscles, not relaxing stomach muscles.
Josh (01:02:31.018)
Mmm.
Josh (01:02:39.306)
Mmm.
You
Josh (01:02:48.329)
Yeah.
Ingmar Jungnickel (01:02:52.686)
There's some things you can do to also change the shape of your body, short of cosmetic surgery.
Josh (01:03:02.536)
goodness this has been just fascinating Ingmar your my brain I feel like is like exploding with ideas and thoughts wow yeah what a what a pleasure so for those listening the your website it's AAiRO.APP got some very cool like flow streamline
Ingmar Jungnickel (01:03:22.189)
Yes.
Josh (01:03:27.834)
simulation model images and some very cool stuff. I am so excited to see where you guys take this in the future, because I will say it, this is the future on multiple fronts. And I'm so thrilled to see somebody doing it, and congrats. I mean, this is just amazing.
Ingmar Jungnickel (01:03:50.444)
Well, thank you. Yes. We're excited about it as well. And yeah, the airflow visualizations you see on the website, everybody that goes and gets an AeroFit session, you do get a report with this 3D viewer and that's where you get your own 3D twin. You can watch the airflow around you. You can move the smoke around. So this is one of these things, this ability to visualize the airflow for you personally and understand how you're interacting with the air.
Josh (01:04:11.028)
So cool.
Ingmar Jungnickel (01:04:19.19)
I think, yeah, checked it out on the website.
Hottie (01:04:22.779)
And also at the website you can find out where a bike fitter or you can find a bike fitter who is AiRO capable and visit them and get your fit done with CFD. Right, that's also on the website anymore.
Ingmar Jungnickel (01:04:37.344)
Yes, exactly. So if you're a rider, we only offer the service through qualified bike fitters and coaches. We do train them up very well. Arrow is only part of the whole equation. We do want people to be comfortable. We do want people to be able to look up the road and not run into trees. We take the right fitter. So if you're a rider, go on our website, find your local AeroFitter. If you're a fitter who doesn't use Aero yet, reach out to us on the website. There's a form.
Josh (01:04:54.442)
Ha
Ingmar Jungnickel (01:05:05.774)
We currently have a bit over 40 fitters worldwide. We're looking to expand internationally, find all the high-end fitters in the world. We're excited about this. It's this win-win for us. The more data we get, the faster we can build better models. So we're pretty interested in getting this out there and in front of people.
Hottie (01:05:27.349)
Well, like Josh said, this has been a mind explosion. Ingmar, thanks for blowing our minds. This has been a great one.
Josh (01:05:30.024)
Yes.
Ingmar Jungnickel (01:05:33.186)
Thank you for having me.
Josh (01:05:34.858)
All right, well, we definitely plan we'll have you back in a year. I want to hear how many simulations we've got going. And love to follow up. right, Ingmar Jungnickel, thank you so much for being on Marginal Gains.
Ingmar Jungnickel (01:05:40.502)
Okay, let's do it.
Ingmar Jungnickel (01:05:46.776)
Thank you.
Hottie (01:05:49.452)
Cool. I'm going to stop.
Leave a comment