Self-Driving Car Technology for People Who Are Blind, How a Swiss Startup Embraces Various Industries: Maël Fabien, biped

Maël Fabien, biped: 0:00

One week after we had the initial idea, we had a 3D printed prototype just out of the hackathon and we were going around Switzerland to let people try it.

Alex, X-Health.show: 0:14

Hi, I'm Alex and welcome to the X-Health. sh ow, where I talk to visionaries behind the latest innovations in healthcare. For the eXtra health of the future. We're in Lausanne, Switzerland, again at Biopôle Life Sciences Campus, big enough that I got confused which one is the right building this time. My guest today is Maël Fabien, a co-founder of biped, who dropped his PhD about speech processing to focus on an AI-powered device to allow people who are blind or with low vision walk independently in places they don't know. biped is a pre-series A startup that developed a smart harness with autonomous driving features that warns about obstacles with 3D beeps. It acts as a co-pilot or a guide, or, as biped puts it, "giving you vision with sounds. You'll hear about autonomous cars technology if technology can replace the white cane building medical devices with customers TikTok and Assassin's Creed yes, the video game, meryl. What do self-driving cars have in common with a device for people who are blind?

Maël Fabien, biped: 1:21

Quite funny, self-driving cars are hitting the streets right, so you start to see them around in the US. They have something built in at the core, you judge them by the quality of the obstacle avoidance, so if they avoid killing someone, you just rate it as a good self-driving system. They're camera-based, so a bunch of cameras all around the car that are able to detect all incoming obstacles and somehow they have an idea of how fast other obstacles are coming in their way.

Maël Fabien, biped: 1:52

I think it is something that, when you look at it from a car perspective, can be also solved at the sidewalk perspective. And this is what we're essentially doing. The problems of a pedestrian are going to be the same. If you're a pedestrian that has trouble in daily mobility, you just basically want to avoid obstacles and you also want to be able to anticipate what is coming your way to be able to avoid it. There's a lot more randomness on the sidewalk than there is on the street but it goes a bit faster in the street. So it's kind of a similar challenge but different criteria to assess how well it works.

Alex, X-Health.show: 2:28

And we're speaking about Honda involved in your project in biped. How did that happen? Because I'm just thinking it's using actually some car technologies in medical device, which I mean, I love that, but I'd love to hear also about how did that happen.

Maël Fabien, biped: 2:52

Yeah, I think it's not a very common partnership that you'd expect.

Alex, X-Health.show: 2:57

It's not, but I'd love to see more of that.

Maël Fabien, biped: 2:59

Yeah, absolutely. It's industry and then on-field applications in medical devices. So, essentially, as we started developing this device for blind and visually impaired people, we started to kind of try to replicate what self-driving cars do from a software level, and so we quickly realized that there are some building blocks of the software that are actually very hard to build if you don't have car data which we didn't have because we're a small company and we did not have years of recorded data that were somewhere on the server to train AI algorithms. So we had to do with what we had, and at some point there were statistical methods to predict where things are going. That works to a certain extent.

Maël Fabien, biped: 3:42

If you see a pedestrian in front of a blind person, for example, we could tell that the pedestrian was somehow coming closer and closer with a certain degree of confidence. But that's something that self-driving cars have solved a while back. So Honda has a research institute, that's their European antenna for their research institute, in Germany around Frankfurt, and they have a strong commitment to sustainable development goals. And they were looking for specific applications of their software that were a bit outside of their usual area of expertise that would be related to self-driving or collision avoidance, driving assist and these kind of things. So they reached out initially, which was super nice.

Alex, X-Health.show: 4:26

I heard, actually, Nils Einecke from the Honda Research Institute speaking at your launch day and I was hearing that they reached out to you so that you can use their technology.

Maël Fabien, biped: 4:41

Absolutely yeah, and they came all the way to Switzerland just for the launch event. That's really one of the coolest things that has happened in the past few months. It's lending this partnership and being able to say that any unit that we sell of biped devices basically runs a piece of software that can run on cars or on obstacle avoidance for two-wheeled vehicles. It's the same technology.

Alex, X-Health.show: 5:06

Okay, so let's go back and let's start from the beginning, from the idea. How did you come up with the idea for biped?

Maël Fabien, biped: 5:15

At the time, I was doing a PhD in speech processing at Idiap Research Institute, which is a lab that is also linked to EPFL in Lausanne, and that was mid-2020 and I was living in Lausanne. I met a blind person crossing the street in Lausanne. Not many people know that, but there is an Ophthalmic hospital in Lausanne that is one of the very good reference centers for low vision treatment, rehabilitation and anything related to assistive technologies to deal with vision loss. And so that person was just going to the Ophthalmic hospital and, funny enough, one of the streets that is close by the hospital is one of the least accessible streets of Lausanne.

Alex, X-Health.show: 6:00

Oh, it happens all the time.

Maël Fabien, biped: 6:03

And so that person was struggling to cross the street and he actually took his phone out and started to do a FaceTime to a friend. And his friend on the other side of the line was basically telling him where obstacles were located, based on what he could see on FaceTime on the other side, which I found to be a fascinating use of FaceTime, but also extremely scary, because the smallest latency could basically create a collision with anything,

Maël Fabien, biped: 6:30

it could be a bicycle, could be low-hanging branch or a traffic sign, a hole in the ground, pretty much anything. And so I just at the moment, had the idea of trying to adapt what self-driving cars are doing – for pedestrians, which is, basically, you use a camera, somewhere on the body, could be in the phone, could be something where you don't have to carry it and use computer revision to detect all incoming obstacles. And then, for the interface, there was still to be determined, how do you convey the information that there is something incoming as a danger. Could it be sounds, vibrations, any haptic feedback, whatever. But yeah, that was the origin story in summer 2020.

Alex, X-Health.show: 7:09

And how did you approach the project?

Maël Fabien, biped: 7:12

So we contacted the Ophthalmic hospital. Alex, X-Health. show: Who's we"

Alex, X-Health.show: 7:16

That was you witnessing the person who was blind.

Maël Fabien, biped: 7:21

So very quickly I joined a hackathon that was organized by my lab. I co-founded biped. At the very beginning, there were three of us, so there was another guy from my office who was also doing speech processing, and Bruno, who became the CTO of the company that I met during the hackathon. And Bruno came all the way from Germany. He came for the hackathon and he was also being interviewed for a PhD position at Idiap lab. And so he was like, if I'm taking the train for 10 hours, I might stay for the hackathon. And that's how we met.

Alex, X-Health.show: 7:55

That's Bruno Vollmer.

Maël Fabien, biped: 7:57

Yes, absolutely. And Bruno was just very good in computer vision in general, and so we had a kind of an instant fit. I'm not a good programmer and Bruno was a good programmer, I liked the business side of things and Bruno, I think, enjoyed the technical side a bit more. And so, yeah, we had the first chats with the Swiss Federation of the Blind and Visually Impaired, with the Ophthalmic hospital, and with SZBLIND, that's also one of the big associations for blind and visually impaired people in Switzerland, and they welcomed the project very much and they just put us in contact with a lot of beta testers at the beginning. So, I think, one week after we had the initial idea, we had a 3D printed prototype just out of the hackathon, and we were just going around Switzerland already trying to let people try whatever we came up with.

Alex, X-Health.show: 8:44

Okay, so my jaw is dropped now. A week? So how did the hackathon look like? I mean, how many hours a day?

Maël Fabien, biped: 8:53

The hackathon was very fun, it was, you know, it was intense, but we did something that... I mean, I would be ashamed of the code that I wrote at the time, right. It was a 3D printed case, a small Raspberry Pi that was for... the computing thing. The hackathon was hosted at the first floor of a building where the ground floor was a hardware shop. So we were spending all of our money on the hardware shop. We didn't find a battery, an external battery to run the thing, so we just bought a 50- meter long extension cable so you could walk up to 50 meters, but that was max, and you had to carry the extension cable in your hands if you really wanted it to work. We just hacked stuff together. I think that was really the spirit.

Maël Fabien, biped: 9:34

None of that was, of course, kept afterwards. At the beginning we did a belt and that's what we tested the week after with the Swiss Federation of the Blind. So basically, we had the first blind person wear the device one week after writing the first lines of code, which was super nice. Week after going to the Ophthalmic Hospital we wanted to show them, how it works in life. And so we took actually an external monitor, like an external screen, just to plug the device to the screen to show them what it did inside. They found it super cool. I was just walking around in the metro with my monitor but yeah... we took it very lightly but we wanted to go super fast and just hack stuff together.

Alex, X-Health.show: 10:17

I love the thing that you went actually straight to the hospital and did not try to think what people who are blind or with low vision can think about that or what would they like. W hat actually did make you go to the hospital? Like, yeah, how were you prepared? You know, this is also about like the company strategy, right?

Maël Fabien, biped: 10:43

A good part of it, I think, is just linked to maybe not knowing what to do exactly at the beginning. Being honest, I didn't know many blind people, I knew that guy that I met in the streets, I didn't have his phone number or anything. So I was like, I don't think I know many people who have vision loss, so I just looked for this kind of reference center and we took the approach of saying: we know

Maël Fabien, biped: 11:12

the biggest risk that we have right now is to do a technology push, so to take technology that we would like to develop and try to apply blindly, if I may, in this market without really thinking about it. And we wanted to avoid that mistake. So we just thought okay, we're a software company, even though it doesn't look like from the outside, but we're a software company, we will have to find a form – maybe it's the phone, maybe it's a belt, maybe it's a harness, maybe it's glasses, whatever – we'll find a form factor that just makes it work. And I think that's the very first meeting that we had with the hospital. That's why we had a monitor, just to show that, okay, this is what's running inside, then you guys have to tell us, based on your experience and your expertise, what do you think people would want, what would they be willing to wear, and then we'll figure out the hardware details around it.

Alex, X-Health.show: 12:20

So what was the goal at the beginning? Or, I don't know, the mission, or what you actually said at the hospital, what you want to do.

Maël Fabien, biped: 12:28

So we promised something we're still working on two years later: crossing the streets safely. The biggest thing we say it was even if you can detect a crosswalk in front of someone using a camera, that is kind of a solved problem. In the low vision community, you still have high risks of being hit by bicycles, other pedestrians, cars, electric cars specifically – they don't produce sound, so it's very tricky for a blind person to be able to detect that.

Alex, X-Health.show: 12:57

May I just give an example here, because actually I just came back from my vacation and there I was crossing a four- lane street on the outskirts, there was traffic lights there and there was actually a person with a cane, assumed a person who is blind, crossing the street and going to the middle of the crossroads and I was thinking in my head, like, "just don't run to that person because you may scare the person. Just observe and see if he's safe. And you know, if he's going like too far to the crossroads, you know, then run. You know basically help or shout or like whatever you know. But is it like where Viper can help?

Maël Fabien, biped: 13:49

I think it's exactly the essence of where biped can help. We basically want to be able to say you can go independently from A to B, A being in a city center, B being in a countryside, for example, if you want to take public transports, if you want to walk on a sidewalk that is quite packed in the city center, hop in the metro, know how many stops you have to do before you have to get off, and then take the staircases that go to the second floor, and then you're on that street, and then you cross that bridge, and then take that bus to go somewhere. It's a very complex task but the vision that we have is to build a stack of software that can basically guide a person from A to B safely and that embeds really two parts: one being the obstacle avoidance and two being the navigation aspect.

Maël Fabien, biped: 14:41

Alex, X-Health. show" So I can imagine that this is actually the vision that the Optalmic Hospital loved, right yeah yeah, because there are very few devices out there that try to do both aspects at the same time. So many of the most of the devices that are currently on the market that can do obstacle avoidance are just obstacle avoidance systems. So it's usually clip- on electronic white canes, so just like a small thing that you clip on your white cane. It uses a dummy sensor like ultrasound or laser sensors, just works super fast, it's rather cheap and it works. But it's been the same technology since the 80s, basically. You take the same sensor like ultrasound sensor, they haven't changed really. They've been embedded on glasses, on necklace, on belts, on white canes and pretty much anything you can think of.

Maël Fabien, biped: 15:36

But it just gives a binary feedback: is there an obstacle or is there nothing? But in terms of obstacle they really think of, you know, if you clip it to a cane, the cane is going to have this movement from left to right, and with this scanning movement, as your cane points to the left, the sensor itself points to the left. So the sensor is going to detect that there is a wall, for example, on your left, and it's going to tell you there is, it's going to vibrate directly to tell you there's an obstacle on your left. And then you start the rotation to the right, it's going to pick up other pedestrians in front of you, for example, there's a front obstacle and then it's going to pick up a set of cars that are parked on the right as another set of obstacles. So, this swiping movement from the left to the right will lead you to. Alex, X-Health. show: To stop, not move. Maël Fabien, biped: Absolutely.

Maël Fabien, biped: 16:24

Completely overwhelming feedback. And you're like, okay, but it looks like I'm completely stuck, right, and it's something that is very hard to learn. So there has been, I would say, mitigated feedback on the adoption of these kind of technologies, even though you could think, in principle, white canes scan floor level elements, these things can scan, and ideally, up to head level. So why doesn't it work better – on the market, not as a technology. And I think this is the essence of having a feedback overload and being able to do only one thing but still like not do it very well, because they cannot filter if you're following a pedestrian, for example, in the crosswalk, it would still warn you about that pedestrian all the time, even though you're walking at the same speed, zero risk of collision. Because it's all thought of in terms of distance rather than in terms of movement and risks of collision.

Maël Fabien, biped: 17:21

And I think that was the first compelling thing. And then the second one that they really enjoyed was probably the add on to the navigation aspect, where GPS are sometimes not precise enough to just say, you know, embed a Google Maps API and then problem solved. Sometimes you need the visual information to be able to be a bit more precise than what the GPS is doing. So if the GPS tells you cross the street and turn right, okay, cross the street, but where's the crosswalk? And this is exactly why you need a camera at some point to be able to detect the crosswalk and that the person follows exactly the middle of the crosswalk, and then it's a matter of ergonomics what type of feedback you provide. But that's very user centric and that can be figured out if you have the technology basis.

Alex, X-Health.show: 18:09

So how did that collaboration start with the Ophthalmic hospital here in Lausanne? What did you start with?

Maël Fabien, biped: 18:17

We first bought a battery so that we would not have to use the extension cable again [laughter]. They had a small committee, an innovation committee, and they agreed to do two things for us. The first one is help us with any resources or guidance, or expertise that we needed to improve, from a theoretical standpoint, the device, to be kind of a premium consultants for us, and that really helped. And the second one was the premises. As stupid as that, but we were testing a medical device. We wanted something that was quite realistic in terms of conditions for these tests. But we had no premises.

Maël Fabien, biped: 19:01

We didn't know what the proper obstacles were like for blind and visually impaired people in the streets and, additionally, we didn't know many blind and visually impaired people. And so they just said okay, let's do just a quick newsletter, we'll say that you guys are organizing this beta testing phase, we'll lend you a room – that was actually like a whole gymnasium, so that was quite big – we'll lend you all the road safety equipments that we have that are the common obstacles that people bump into, traffic signs, plastic branches to simulate trees, for example, in the street and this kind of things.

Alex, X-Health.show: 19:37

So this is how they train people to use canes, right?

Maël Fabien, biped: 19:40

At the very beginning, yes, and then they quickly go out in the streets.

Maël Fabien, biped: 19:44

But at the very beginning the training is usually in two steps.

Maël Fabien, biped: 19:47

First you have to learn to walk without a cane, so it's really hand based and understanding your surroundings also based on sounds, and then they start using the cane, first indoors and then they start to go outdoors. And we actually learned the existence of the job of a mobility trainer at the time.

Maël Fabien, biped: 20:06

It's not a very common job, but there are... at least we've met about 30 mobility trainers in Switzerland and they teach blind and visually impaired people to discover methods to be able to navigate in their surroundings, ranging from no white cane to white cane to other kind of assistive technologies. And they have very deep expertise on how you teach a blind person to go from A to B. Because sometimes if you're blind and you want to just go to another supermarket, for example, that you've never been to, if you don't have a relative that can help you and if you don't have the methods that are completely acquired to do this mental mapping for yourself, then you need external support from a mobility trainer so that they teach you how to go to this new supermarket.

Alex, X-Health.show: 21:18

It actually reminds me about a person that I met a long time ago. He was blind and he was so... nonchalant, walking the street. I went to meet him and I actually witnessed him approaching our meeting place, his cane under his armpit, you know, a cigarette in his mouth – he owned the street. But then I thought, that's the place he knows, that's the street he walks every day, for five to ten years. And the problem is when a person who is blind goes to a new place, right, because then, like you say, there might be a tree branch, or everything is so new.

Maël Fabien, biped: 22:13

Yeah, absolutely, and I think that the job of mobility trainers is actually very interesting because it's just about adding dimensions to the lives of blind, visually impaired people in terms of spatial exploration Is really range from that linear thing: I have my place to maybe workplace or grocery to adding multiple dimensions. And the ultimate thing is usually you teach techniques that once people master them properly they can just replicate that by themselves in unknown places. And that's a bit the dream all the companies chase, trying to chase devices that sell some kind of independence, like what we're doing in terms of assistive technologies. W e're all chasing this, this thing where basically someone would be able, would have a meaningful feedback to securely explore new places conjointly with the kind of techniques that people have learned in mobility training. And I think, one of the things that has very often been criticized in our space, is that some companies go there with very bold, statements saying, you can drop the cane, it's the only assistive technology you'll ever need, etc., etc. And then they fail at getting adoption from orientation and mobility trainers, which are usually the prescribers or they're usually the ones recommending this kind of devices for their blind, visually impaired patients, so kind of trying to bypass this.

Maël Fabien, biped: 23:45

This part of the chain somehow leads to too much of a promise on the technological level and being in it, I know it's not ready to say, "you can drop the cane and we have figured everything out. We need for people to be able to just move freely, and the biggest bottleneck is not exactly where you think. Right, it's not so much on the camera or like vision, understanding or this kind of things. It's more on the communication of the feedback. Because I think there are lots of cool things happening in neuroscience, but just the auditory feedback, just giving beep beep beep-beep-beep as the obstacle is coming closer, which is what we're doing, which is what car manufacturers are doing when you park your car. It's just a common thing, but it's very little amount of information being conveyed at a time. We add a 3D effect to it, so you can also differentiate where it's coming from. The sound is being played more and more often as obstacles are coming faster, for example, and then we can change the pitch of the sound if it's a head level or a ground level obstacle, but it's still on the tiny bit of information conveyed every second versus a visual information that is...

Maël Fabien, biped: 24:53

I remembered that from... I think a data visualization class that I had, that from a visual standpoint, they say that you could process kind of... I am trying to remember these numbers, but about 30 megabits of data per second, whereas on the audio channels, if you only convey sounds, people would be able to process – it was in a matter of kilo bits per second. So you have just gigantic factors between both, and focusing only on one channel is probably not the right way, and I think there are good progress being made in the neuroscience field that kind of make it hope that we'll be able to bypass the auditory feedback and somehow, with the brain stimulation, be able to provide specific type of impulse that can give some kind of notion of shape or convey kind of 3D shapes. I think that that's super interesting. We're probably 10-15 years away from that thing happening.

Maël Fabien, biped: 25:49

Alex, X-Health. show: So something like projecting on the brain or in the brain, or, as you're saying, stimulating the brain to project something. Maël Fabien, biped: If you look at what companies like Neura link are doing, they have identified... I mean, there are many and I'm very uneducated in this space, so I won't have too much into this but at least I'm following what these guys are doing. And they have identified regions in the brain that are responsible for specific parts of the perception, as you know, pretty much like pixels, and from the latest video they published, it looked like they were able to rebuild an image with a decent resolution, with a notion of depth, and so people would be able to understand where shapes were located.

Maël Fabien, biped: 26:37

Alex, X-Health. show: For people who couldn't see at all? Maël Fabien, biped: They're doing animal testing right now. At least the promise is that you could somehow get a person that has low vision that cannot be treated, where you could identify which region of the brain was responsible for specific parts of the perception, and then, if you have a camera on the outside, you could actually bypass any kind of human physiological channel for the feedback and directly plug in the specific region. And there are lots of concerns around, what happens if you've been blind for more than nah-nah-nah, have you

Maël Fabien, biped: 27:19

lost some kind of plasticity. Alex, X-Health. show: Exactly, the brain plasticity is what I wanted to mention. Maël Fabien, biped: And the resolution you can get and the adoption of these kind of technologies. And maybe this will never go through any ethics committee or whatever, I don't know, right, I watch it from the outside because it's way beyond my reach and I'm not smart enough to dive into this, but it's just super cool and super exciting to think that somehow, by, like, 2050, it might be a solved problem.

Alex, X-Health.show: 27:48

People are working on it at least, right, and it's really amazing. Let's go back to how many people signed up for your tests. Maël Fabien, biped: That was quite a quite a bit we got.

Maël Fabien, biped: 28:03

I would say, we woke up some day with this newsletter article published and probably about twenty or thirty people who wanted to test it.

Alex, X-Health.show: 28:13

That's not bad, is it?

Maël Fabien, biped: 28:14

I mean, I went from knowing one blind person to knowing thirty of them. So that was quite fun. We had this room and we had these big boxes, all the traffic signs, and so we just organized the first test.

Maël Fabien, biped: 28:36

That's when the device was not a belt anymore because we quickly realized that people were actually walking with their white cane and the technique is actually, bring your hands at the middle of your body and then that's where you start the rotation, the swiping of the white cane, from. The hand was exactly in front of the camera. So we started to move the camera up and we tested a GoPro harness at the time that we clipped on the body and our small camera module that we plugged on there. That was the first session organized and we had two feedbacks. The first one was, it's way too hard to clip or adjust because GoPro designed this stuff for guys that are doing snowboard or whatever, but they take the time to really clip and adjust, and they can see what they're doing. And the second one was, it's producing way too many sounds. So there was a bit of a cold shower for us at the beginning because it was not the perfect adoption from the start, like yeah, you guys rock, it's perfect. It was more, try to rework this stuff and we'll see.

Maël Fabien, biped: 29:47

And so we reworked the hardware, we hired one hardware intern and then, from the software standpoint, that was the very beginning of the moment where, I mean we're still working on this stuff, but at another level now,

Maël Fabien, biped: 30:04

how do you reduce the amount of sounds, that the device is generating, to only focus on the moments that really matter the most? And that was quite hard. But this is exactly what Honda is bringing to the table because having a notion of how things move around is the perfect way to say, ok, that thing is actually going a tiny bit faster than me, no need to have a sound played for that. But on the other hand, if that thing has a tiny risk of collision with my shoulder, then I should warn the person about that, because you'd be judged by the quality of the obstacle detector. If you're putting lives in danger during a test, some people like we've heard stories of in previous tests of these kind of technologies, people were just dropping out of the test in the middle, saying, I just collided with something and it's your job, guys, not to make me collide with something, I quit. And so some people could eventually leave the kind of beta testing program very easily.

Alex, X-Health.show: 31:02

I remember Nils speaking at your launch event that actually they worked for some time on this machine learning, he mentioned 20 years. 20 years ago they started this machine learning and collecting the data to learn how to detect the obstacles that might be dangerous and that are not, as you're saying, moving with you. So like, say, pedestrians moving with you or moving across the road, or a car standing there. Without that, where do you think you'd be with, with biped?

Maël Fabien, biped: 31:47

I think we'd be a bit closer to the value proposition that other companies might have, which is this binary feedback, is there something or not?

Maël Fabien, biped: 31:57

But if you don't have a notion of movement, properly speaking, anytime there is something that is closer than, say, three meters, I just play a sound. That's a bit restrictive as a hypothesis and I think this is really what we're bringing on the table. I still think that could work, right, we had a base model for movement estimation. But it was the kind of a model that is not precise enough to properly bring something and it created probably too many false positive in a sense that you have too many things that are predicted to come in your direction, because the estimations of the cameras sometimes are not perfect. I think it's really bringing these two aspects of technology readiness, because they've been working on this extensively, and also the name, because first it's a great partnership and it's an access to a portfolio of other algorithms for further collaborations also.

Alex, X-Health.show: 32:59

[MID-ROLL BREAK] This episode is brought to you by the X-Health. s how and me, so if you like this podcast, be generous, hit Follow, leave a review. That will help me invite more such amazing guests for you, thank you. [MID-ROLL BREAK ENDS] How many prototypes did you develop?

Maël Fabien, biped: 33:17

We did six different prototypes, so it was like six times restarting the hardware completely. What were the changes? The first one was obviously this belt and then we added the battery. The second one was already a GoPro harness and we had some kind of a box in which we had all the cameras and the computing, and the battery, and everything, and we still had a jack cable because you could only listen to audio feedback not via bluetooth but just via wired headphones.

Maël Fabien, biped: 33:50

And then we redesigned this case so that it would fit closer to the body and at some point we just realized, okay, we have to drop the GoPro thing. It's not going to work out. It's a cool support, but it's not going to work out for end users. And so that's when we started designing the kind of vest format. It's kind of a harness where you put it around your neck. There are two straps, one on the left, one on the right, two parts that go on your chest. On the left side of your chest is going to be a set of cameras, on the right side of the chest it's going to be a small computing unit and to balance the weight behind the neck there's a battery. That was the beginning of the first form factor where you didn't have to adjust anything, which was quite interesting, and we got much better feedback as we did these adjustments.

Alex, X-Health.show: 34:41

So you just put it on and you basically start walking.

Maël Fabien, biped: 34:45

Yeah, and there's an ON button and it can start directly or we have a smartphone application, so that's something blind and visually impaired people will control over voiceover. They have also voice feedback whenever they press a button and they can change settings from there. The settings might range from, in which situations do you want to get a warning? Do you want to detect holes? Do you want to detect things a couple of meters ahead or just way closer to you? They can do an audio training because we also realized that something that seemed crystal clear for us to start using required lots of training in practice.

Alex, X-Health.show: 35:24

Could we hear it? Because we have the harness here. Could we hear it beeping?

Maël Fabien, biped: 35:28

I can make you listen to the sound because on the smartphone app I have a sound recap. It basically plays the whole sound set that we have in case you are somehow lost, like, what is that sound that I just heard? It's not going to have the stereo effect, the left- right balance, but this is basically what it feels like when an obstacle is coming closer when it's moving fast. So the sounds are being played faster. biped Speaker: When it's on the left.

Speaker: 36:05

When it's on the right. When it's a head level obstacle.

Alex, X-Health.show: 36:15

Okay, they change a bit, right? The tone changes.

Maël Fabien, biped: 36:18

Absolutely. That's how we got into sound design, which by itself is a complete different field that we had no idea even existed. We wanted to play an immersive experience so that people could really listen to these sounds. So part of the job is to people don't have to listen to many of these sounds but the counterpart is that they should be able to live the biped experience for hours straight, so they should be able to support these sounds for two hours or whatever. And we just make sure there are not too many moments where these sounds are actually being played.

Maël Fabien, biped: 36:54

We contacted audio programmers and people who were just very good at designing audio experiences. We've found two guys who worked on Assassin's Creed and they were just very good at what they were doing, and they had this experience from the gaming industry, from very immersive sounds, sounds that were a bit like round and that you could really listen to for four hours but at the same time they would convey the right amount of information when it's needed. They basically told us, guys, if you want to reflect elevation ideas, you need to play with a pitch of the sound. It's going to make it feel like it's coming from a head level or a ground level. We can start triggering the sounds when the obstacle is that far and then we're going to ramp up the volume and then play it more and more frequently. They basically took what they knew from the gaming industry and applied it for us.

Alex, X-Health.show: 37:47

Okay, so it's basically Honda meets Assassin's Creed meets visually impaired people or people who are blind.

Maël Fabien, biped: 37:57

Yeah, it seems like a bad movie crossover.

Alex, X-Health.show: 38:02

It sounds amazing, really. Okay, so how did you w ork on iterations? When I looked at your website, there are still some to-dos on your website. You launched the product knowing that there are some jobs to do, right? So what did you accomplish? What do you still need to do and why did you launch the product that didn't have all the features that you imagined or you wanted to?

Maël Fabien, biped: 38:36

I think we tried to have a lean approach. We tried to embrace the notion of launching, not launching too late. In that sense, we wanted to launch as early as possible and we kind of wrote down what our MVP would be like and for us our MVP was we need to be able to update the system.

Alex, X-Health.show: 38:56

That's minimum value proposition.

Maël Fabien, biped: 38:58

We need to be able to update the system remotely because obviously we'll not be ready. So criterion number one: updated remotely. Second, it should not break. So the hardware should be stable enough that we don't lose access to a camera or that if it falls people can keep using it. And third, there needed to be at least one base feature. That was there, and so the base feature for us was a beep if there's an obstacle. We kind of worked out the very fundamentals of the value proposition. That was it. And when we got everything together, that's exactly when we started launching. And that was way too early. That was right before the Consumer Electronics Show in Vegas, where we had the opportunity to do the official launch in the US, and there was a cool momentum.

Alex, X-Health.show: 39:48

When was that exactly?

Maël Fabien, biped: 39:51

The launch event was December 2022. And we just flew at the very end of December for the show that took place the first week of January 2023. And, yeah, that was a bit jumping in the cold bath. We realized that most of the stuff we thought worked on the MVP actually needed a lot of rework. But that was just cool learnings, I think, along the way. And we wanted to be transparent with end users and the fact that they were buying a prototype, or at least a beta release in that sense, and then the software would evolve.

Maël Fabien, biped: 40:30

From the hardware standpoint, we basically try to freeze as much as possible, except how the things are organized inside. We just make easier and cheaper to produce now, but it looks the same for end users and any software update that we do is also compatible with the previous devices. So there was a minimal risk at this level. We had the hardware readiness, we could update it and we had the base beep sound being generated, which was good enough for us to launch. Of course, what it was missing was handling complex scenarios. So deeper integration of whatever we're doing with Honda, deeper understanding of very complex and challenging scenarios, so that could be, you know, weather conditions or kind of indoor navigation aspects. That's something we've really worked on to improve the obstacle experience. And the second bit is what we started working on, which is the navigation aspect. So really trying to be able to know where am I located, where do I go next, and get these turn by turn instructions. We have the first features of this that are out there.

Maël Fabien, biped: 41:43

But basically we have an open roadmap on our website where end users just go and then refresh. They get a newsletter update on, hey, here are the new things to test. Could you let us know what you think? And as soon as you try the device, for example, on the app, there is a pop up that comes up and just asks for your feedback, where you can just drop a voice note on WhatsApp, for example. So they all have my WhatsApp number and they just keep providing feedback that way. And I think now we built a small pool of very frequent end users who just debug most of the stuff for us and they let us know if something's better or worse or if it's broken or if it works as intended, and that's just super cool. Alex, X-Health. show:" o, that's amazing.

Alex, X-Health.show: 42:24

And who are these early adopters? Or at the beginning, prototype adopters really?

Maël Fabien, biped: 42:34

So we're selling D2C right now, Direct to Consumer, meaning these people found us via podcasts, fares, newsletter, whatever.

Alex, X-Health.show: 42:47

And they were just straight saying, like, I want this.

Maël Fabien, biped: 42:50

And they just placed an order on the website. It still blows my mind but some people would do that and that was the moment when we realized that could be something. I won't call it product- market fit, it's still too early, but if that keeps going that way I think at some point we will be able to call it that. W e always had the assumption that most of them would start the device for the first time and do the setup with a relative who would help. It was not the case at all.

Alex, X-Health.show: 43:28

Why would they?

Maël Fabien, biped: 43:29

Yeah it's, you're selling a device for independence, so it should be easy to set it up independently. And at the very beginning we, for example, just put a piece of paper with a QR code to download the app. And one person was like, yeah, but I haven't found the QR code. And we said, yeah, it's on the paper. Yeah, but if there is no tactile overlay, I can't know there is an app, so can you just send me the link? There's simple mistakes that we did like this, but that were solved very quickly. And now there is a an experience where you can set it up completely independently.

Maël Fabien, biped: 44:05

These people are, some of them, working actively, they work in accessibility departments, they work in IT, some of them are directly involved in end user associations, some of them are retired, so kind of varied. But we just realized very quickly that our testers are quite spread out. Very early that got to a level where you had to sport multiple languages and do maintenance in various countries, which is good challenge, because then if you succeed doing that and you add the scaling factory to it, then you're kind of international by culture in the company. Alex, X-Health. show: I read there was 250 testers.

Alex, X-Health.show: 44:55

Does it include these early adopters?

Maël Fabien, biped: 44:58

So that was before these early adopters. A long the six prototypes we developed, every time we tried to have five or six common testers but we also wanted to add new people to the beta testing program and get their feedback on other aspects of the device. And also because we anticipated that we wouldn't be able to sell all of our devices in Lausanne, because that would be a very niche strategy for now, we wanted to get feedback from various places. So we tested the device in Paris, in London, in Amsterdam, in the US, in various places, sometimes in complete countryside, sometimes in busy city centers, just to get feedback.

Maël Fabien, biped: 45:40

And as a program kept rolling, some end users didn't know that, but when they contacted us... For example, we had one guy in Long Beach and I was like, I have something planned in the US later that year, I'll just go. He probably won't know I took a six hours flight just to meet him, but that will be very interesting to get his feedback on his use case around that. And then we tried to book one or two more beta testers because there would just be burning too much money otherwise. But yeah, we tried to have an approach where we embedded as many people in there as possible.

Alex, X-Health.show: 46:12

That's customer- centricity level high. I love that, totally love that. So you didn't go the medical device certification road, or are you going to catch up on this one?

Maël Fabien, biped: 46:25

So we did. We are a class 1 medical device. The more recent changes in Medical Device Regulation obstacle avoidance systems were not class 1 medical devices, but then all white cane manufacturers also had to catch up with this and do the regulation in the last couple of years. It's still the easy part of medical devices in the sense that it's self declared and t he regulatory constraints are actually a lot easier. Alex, X-Health. show" he pricing strategy.

Alex, X-Health.show: 46:57

– could you tell a bit more about that?

Maël Fabien, biped: 47:00

We're doing a subscription which in hardware is not that common. You start to see more and more of that. We do 129 Swiss francs a month, still quite expensive, but that just allows people to spread out the payments a lot. After three years people get to keep the device and the subscription stops. In the meantime, if they don't want to keep it, they just return it and the payment stops.

Alex, X-Health.show: 47:25

After three years they will still get the updates on the app?

Maël Fabien, biped: 47:29

They get the updates for three years and then we'll see.

Maël Fabien, biped: 47:32

The company is not even three years old, but the what we anticipate is that we'll just freeze whatever last version they were running and we just ensure that compatibility with the app. The idea behind that was, on one end, we could try to go for the reimbursement of the solution, because if we charge three years of subscription upfront, we are selling it for 4600 Swiss francs approximately, but that actually is a very steep price when it comes to just doing one down payment.

Maël Fabien, biped: 48:03

So we could do that at a discount for early adopters to join, because all the features are not exactly there, we launched early, etc., etc. The subscription system was a bit more lean in the sense that if we're really a software company, we're assembling hardware but we're just selling the updates of the software. And I think it's just aligned us with the interests of the customer in the sense that right now what we're doing is we're tracking usage of the device from end users and then when we see that people use a bit less or that there is a bad adoption, for example, of a certain feature or release or whatever, then we work on it just like a software product would do.

Alex, X-Health.show: 48:46

So like, say, iPhone update.

Maël Fabien, biped: 48:50

Yeah, yeah, absolutely.

Alex, X-Health.show: 48:52

All right, and how did you actually approach funding for the company? Just before we started recording, you told me that you actually approached it with self funding for the start. But I saw you pitching, and actually that was your last and winning pitch at the De Vigier prize here in Switzerland, in a lovely summer house, castle, in Solothurn. So you did some fundraising, and quite successfully so.

Maël Fabien, biped: 49:29

Yeah, to a certain extent we can call that successful, even though I have my reserves... I could have done things better. But we're always smarter the day after. The way we approach that was we incorporated the company at the beginning of 2021 and we raised our first round late 2021. So we had a first year where we knew we had to de-risk the product and we knew it was simple enough to build, that there would be some kind of expectations from investors, that we have at least something to show,

Maël Fabien, biped: 50:06

it's not the kind of project where you can just raise on paper or at least we didn't have the right profiles to do so, and at the same time, complex enough hardware medtech niche markets. Kind of not best green flags you can think of. And so we self funded. At the beginning I was doing lots of freelance missions and at some point I had a freelance mission that came up and that was just way too big for me to take. That was like the equivalent of three months of full time work. But as a freelance you get to charge, also on a daily rate. So this decent amount of money could actually live the whole year for the company.

Maël Fabien, biped: 50:52

I was about to decline it when my co-founder was like, you know, we could actually take it and do it together and that would provide funding for year one, basically, which is what we did. And then, after the end of the year, we raised a pre- seed round that was mostly convertible loans. At this stage, 2022 was really focused on getting as many grants and non-dilutive funding as possible, and 2023 was our seed and trying to plan where we go next, basically.

Alex, X-Health.show: 51:29

Right, and then, I saw your TikTok. So just wanted to say congratulations to be on TikTok. You're the first startup I'm speaking on the podcast that actually is on TikTok, and I think rightfully so, because I think your audience is there. Did you check if they are there?

Maël Fabien, biped: 51:49

We probably don't have the most straightforward use case or like customer base to be on TikTok, but we did a bit of market research. We took it from the approach where I was like, I don't have a personal TikTok account, or I mean now I do, but I didn't have a personal TikTok account. I was really kind of afraid of social media because, for me, it was a big time trap. Alex, X-Health. show: Well, it is.

Maël Fabien, biped: 52:16

Maël Fabien, biped: Yeah, and I wanted to approach it from a professional level.

Maël Fabien, biped: 52:19

So we hired someone who's doing social media strategy and we started creating content with that person directly and that helped a lot structure what we wanted to do as a company. We defined who our communication targets would be on each social media and on TikTok there is blind community, usually under the #blind talk and there are also lots of occupational therapists and mobility trainers and so, funny enough, they just share educational content about their job. There are also blind and visually impaired influencers and some of them have millions of subscribers because they just share about their life or how they go around, like random stuff in life. We found a small community there because most of it's funny, because it's one of the features of TikTok, but it has this voice description also t hat people are just very much used to put on these videos, and so that was kind of a fit, even though I didn't expect it at all at the beginning.

Alex, X-Health.show: 53:25

You can go there and just find answers to the questions you might have about biped. But I mean, we were speaking about funding and actually I found, in November 2022, you just posted that you, well, I'd say, you didn't succeed – and I translate what you really said there with your first VC pitch. So what did you learn? How did it look like?

Maël Fabien, biped: 53:55

You've probably understood it, I like to be very open about the things we f... up and that was a good one. So the very first pitch that we had at the time I was still a PhD student before I dropped out, and so we just booked a meeting room inside the research institute I was in. I just asked them whether I could bring an investor there and they were like, I don't see a reason why not. So, you know, I book it. And I had met the guy before and he was like, I'd like to learn more, can we have a deeper chat? And I didn't prepare like anything specific.

Maël Fabien, biped: 54:31

I had my eight- slide deck, I tried to keep it as short as possible. And then when he came back, I just read the eight slides and I thought he wants to meet and he wants to... probably not go too much deeper into it, and he just wants to meet the other two guys and that will be enough to give us money.

Maël Fabien, biped: 54:48

And... Alex, X-Health. show: It wasn't. Maël Fabien, biped: And it was not right, it was not! [laughter] And I was kind of clueless, because he asked lots of financial questions, just the simple things... I mean, that was a stage we had prototype that sort of looked like the final device, and I just missed lots of financial knowledge to be able to answer questions like what is the cost of production of the device and how many units do you plan to sell in 2023? And: what is your cost of acquisition for one customer? Simple things like this that now I have on top of my mind all the time. But I just didn't have that.

Maël Fabien, biped: 55:28

And that was a good training in the sense that when he left, I was really pumped. I was pretty sure we were going to get an investment. T here was one lady who was also with him and she wrote me an email saying that it would be good that next time we have an investor pitch, we actually get ready to be mindful of everyone's time, because this is one of the worst pitches she has seen in the past few months.

Alex, X-Health.show: 55:57

But so lovely of her to give you feedback actually.

Maël Fabien, biped: 56:00

Yeah, I mean, the guy was also a bit weird and reconsidering it, I don't think it would have been a good fit in the investor pool, but that was, I think, a hard way to learn. But I think our next pitch is one bit better.

Alex, X-Health.show: 56:13

OK, so let's look into the future, what's coming next?

Maël Fabien, biped: 56:22

So two big developments that we're doing. The first one is we want to have some more features on the market of blind and vision impaired people, so really live by this promise of doing the navigation and getting that solved. The other is how do we take the software that we develop and embed it in other form factors that can answer other use cases and the challenges that blind people meet? People who have a lighter visual impairment also meet it but people who are elderly and struggle maybe in busy city centers might also face it. And so what we're building as a vision for the company is the fact that to be able to get insights on how to navigate better, you don't really have to be blind, because what we're dealing with is is not vision loss. We're dealing with the fact that you have the ability to really focus on an obstacle that might be coming around you.

Alex, X-Health.show: 57:25

And you mentioned the electric cars. I mean, I don't notice them and I don't have visual impairment, right.

Maël Fabien, biped: 57:31

Yeah, so, and we're starting to get more and more feedback from sighted people who comment on this stuff. And I think there is a path towards, needs to be a lot smaller, towards a very mass market kind of product, a couple of years from now, definitely. But in the middle there are lots of use cases that I think are just going to be super useful for us to focus on. Visually impaired and elderly people as a first expansion of what we're doing right now, which is more focused specifically for blind people. And as hardware evolves and as we get more funding and more means to put into this kind of device, then more skills to actually also be able to make that hardware work.

Alex, X-Health.show: 58:19

So how do you want it to make it happen? Or what do you need to make it happen?

Maël Fabien, biped: 58:26

In terms of needs, we'll take it step by step. We're very much focused on the execution and try to live by the numbers that we also share with our investors. The core thing is in the execution. We want to go from doing a device only for blind people to expanding to visually impaired people. That will need a hardware rework and that will need probably a team of 10 to 12 people and somehow, whatever way we budget that, we're anywhere between 3-4 million. S omething we're working on next is also getting the good momentum with our seed round that we can actually convert a good part of it into sales that also demonstrate the ability to execute and then to live by pre- series A or however we want to call the next round.

Alex, X-Health.show: 59:15

OK, so when would you open the first round?

Maël Fabien, biped: 59:18

That would be probably open towards the end of the year.

Alex, X-Health.show: 59:21

So end of 2023. Ok, so wishing you, well, success in that.

Maël Fabien, biped: 59:27

Yeah, thank you.

Alex, X-Health.show: 59:28

Thank you very much for the conversation, Maël.

Maël Fabien, biped: 59:30

Thank you so much.

Alex, X-Health.show: 59:31

And for you guys, if you wanna see what's new at biped, please go to biped. ai. Thanks again. Maël Fabien, biped: Thanks. [POST-ROLL STARTS] I'm totally impressed by the audacity of researchers turned startup founders, doctors turned entrepreneurs or ordinary parents turned healthcare innovators. People battling the battles that no one fought before. For the eXtra health of the future. So if you see a startup posting on LinkedIn, show them some love, hit Like, comment, That's fabulous. If you have a couple of drops more of that altruism, follow the X-Health. show, leave a review here. I'll be able to bring more of these visionaries to you. So a big thank you. You're awesome. See you next week.

Speaker: 1:00:16

The information in this podcast is for informational purposes only and should not be considered medical advice. If you have any medical questions, please consult your healthcare practitioner. The opinions on the show are Alex's or her guests. The podcast does not make any responsibility or warranties about guests' statements or credibility. While the podcast makes every effort to ensure that the information shared is accurate, please let us know if you have any comments, suggestions or corrections.