In this episode, we talk to Felix Laumann, founder of Neural Space, a cutting-edge AI platform that is focused on breaking down language barriers. We hear their founding journey and learn how their text and speech APIs are opening up local languages all around the world.
Episode Links
Connect with Felix on LinkedIn
Episode Transcript
Matthew Todd
Hi. My name is Matthew Todd and welcome to Inside the ScaleUp. This is the podcast for founders executives in tech, looking to make an impact and learn from their peers within the tech business, we lift the lid on tech businesses, interviewing leaders and following their journey from startup to scale up and beyond covering everything from developing product market fit, funding and fundraising models to value proposition structure and growth marketing.
We learn from their journey so that you can understand how they really work, the failures, the successes, the lessons along the way, so that you can take their learnings and apply them within your own startup or scale up and join the ever-growing list of high growth UK SaaS businesses. Welcome back to the podcast joined today by Felix Laumann from NeuralSpace. Felix, great to have you here today.
Felix Laumann
Thank you very much for inviting me.
Matthew Todd
No worries, I’m looking forward to the conversation to kick things off, can you tell us what NeuralSpace? What is it that you do?
Felix Laumann
So NeuralSpace is an NLP that means Natural Language Processing company, we work on both text and voice and they automate all kinds of different things. So the all of us Google Translate, most likely, that’s one of the tasks that are that fall under NLP. But there are also many more, for example, voice transcription, when you automatically generate subtitles to a meeting, or lots of servers can already do right now.
But also many more, like what is called transliteration, translating between different alphabets, automatically extracting keywords from text, or even finding the meaning out of a text snippet. That could be for example, used in a chatbot. All of that we do in more than 100 languages and our speciality is in this locally spoken languages in Asia, Middle East and Africa.
Matthew Todd
Okay. We’ll get into more details about how the platform works, I’m sure but why go after those localized languages and, what led you to try and create this in the first place?
Felix Laumann
There are really not that many solutions to support these languages. That’s number one. I think, about a year ago, also Microsoft and Google moved into into these languages and they all say now they have more than 100 languages in the offering. However, it comes down to how accurate your models are for those languages, right? So we have all used Google Translate, or we’ve all use some kind of voice transcription service, we always see they are never 100% accurate. These mistakes are already present in English. But the more local you go, the more mistakes are generally are. So you really have a huge difference between an offering by NeuralSpace, which makes maybe 10%, or 8%, of all words, as a mistake.
But then you have an offering by Google or Microsoft making 30% of all worlds being, for example, wrongly transcribed, and 30% is a huge number, right? Because very often, it’s a the post edited by a human that obviously requires less time when you get it more accurate in the first place. But very often, it’s it’s also really like vital information, right? Or vital services. So vendors, for example, a chatbot, that allows you to support to book an appointment for your COVID vaccination. You want to get 99%? Correct, right? So really, the accuracy matters a lot in those languages.
Matthew Todd
What was it that led you to try and tackle this kind of problems in the first place? Was it something that you personally experienced or just an opportunity that you saw?
Felix Laumann
Two really big influences, I would call them or really big realizations. So first, was the technical challenge. So you generally deal with a lot less data than what you have in English, or let’s say German, and French and Spanish. You have 1% of data in let’s say, Arabic dialect. We only have only 1% of data than what you would have in English. So it’s a it’s a technical challenge, which makes it very interesting.
But secondly, is the impact that we can have when we provide adequate solutions in those languages. I was lucky enough to travel to lots of countries about one of his locally, locally spoken languages are spoken and so very often the people do have a smartphone, and they do have internet access. But they’re very often forced to use it in English, although they really are not comfortable English, or they really barely speak it.
So the keypad was in English, still when they wanted to access like WhatsApp, the WhatsApp interface, right? It’s never in a local language. So lots of things I saw were wow, okay, these people have access to technology. But then they used to, they’re forced to use an English or French in some countries. So I was thinking, okay, but when we provide that local language support, we can really make our lives a lot easier, and hopefully even make them adopting technology even faster.
Matthew Todd
So how long was it then between spotting that problem, and validating it having something that you were convinced would provide a solution to some of those problems?
Felix Laumann
I would say, one and a half years to two years. It was because we needed to build the technology, from scratch, really, there was nothing existing, we had some academic resource. It really took some good effort to get it all into, let’s say, product by the shape, right? It actually took one and a half to two years to really make a product out of it.
Matthew Todd
Was that a single language you were targeting at that point? Or were there multiple languages you were looking at?
Felix Laumann
We were looking at Indian languages, and about 11 locally spoken Indian languages that we tested in the beginning. That was partly because we had early market traction, or let’s call it interest from companies in India, my co founder was Indian. So it was a bit easier for him to actually understand or human validate what our models are predicting.
We said, okay, let’s tackle India. There was 11 local Indian languages and that expanded over time. So it takes us about, I would say, three months to add a group of let’s say, 20 more languages, 15 to 20 more languages, we normally tackle them in groups. Very rarely only add one new language.
Matthew Todd
Were those initial paying customers from your existing network in your founders network?
Felix Laumann
Yeah. So it is a bit of a leap of faith. It’s still very risky to buy from the further, because who knows? Like maybe in the year, they’re not in business anymore, then you need to rewrite your entire technology stack and all of that. So it was my personal network and quite a good license for them. Let’s call it that way.
Matthew Todd
Yeah, I see. And how many different languages are you supporting and what kind of regions are you now working in?
Felix Laumann
There are nearly 100 languages that we have actually now. India still very much, very much one of our most attractive markets. It’s just a huge language. They also have diversity. So it’s not only different dialects, they even use different alphabet and it’s obviously very difficult to build accurate technology for those.
Secondly, the Middle East, we have seven different Arabic dialects and are offering right now. It’s also market more people where people generally mix languages. So that’s always a very interesting for us and other people do that in other countries. But a mix Arabic with English or in Morocco, Tunisia they mix Arabic with French. So that’s obviously very difficult for especially for the voice technology to transcribe that accurately.
We are about to release it in two weeks the very first models that are out in the market that can accurately transcribe these code mixed or code switching languages.
Matthew Todd
In terms of commercializing the technology. Did you look at providing products that you would license access to, like a SaaS platform of some kind? Or are you providing access to the technology via API’s, etc. How have you approached opening that platform up and commercializing it.
Felix Laumann
Looking at what other big tech players doing and building up that model more or less. We charge per API icon, or per minute of speech? When it’s kind of a permanent API, let’s put it that way, when it’s a streaming API that is always running. We just measure how many minutes of speech we actually transcribing, or how many minutes of speech we are generating, with a syntetic voice.
That has been kind of what the market has been being used to, we have had some requests that people asked just for license, so unlimited usage over one year. As a startup, we are quite flexible, right? So we are not, not that quick, not at associated with a very strong processes that cannot be changed. We are generally very open to what works best for the customer. But the pro API pays a per PI base is what has the big tech adapted and not say how they sell. We just wanted to have the least change for our customer.
Matthew Todd
Speaking of the the big tech companies, then of Google and Microsoft, it seems like you’re very much starting out with some of the biggest tech companies in the world as a potential competitor. How do you? How do you feel about that? How do you approach trying to compete with those companies?
Felix Laumann
To be totally honest with you, in the beginning, I did not think it’s that big of an obstacle. So it’s just incredible with what kind of credibility and trust these companies come right. So both these big tech players enjoy a huge amount of credibility in the market and that’s like totally understandable. Where we differentiate as a startup and I think that’s not only applicable to NeuralSpace, is providing the personal support, and being quite flexible in what you offer, right?
So for example, we have had a few requests by customers to provide on premise installations of the technology. The big tech providers will most likely not do that. There’s really something that you can offer as a startup or to big tech that big tech can’t provide. Also being really flexible with your processes. So when a customer wants to have an unlimited license, it’s something we can do, right? It doesn’t need to have a five layers of approval by anyone. It’s just like what you can offer.
Matthew Todd
In terms of finding customers, so we’ve had some other people on the podcast founding companies that rather than providing a service that consumers whether it’s b2b or b2c that we use directly, they provide an API to allow access to their technology. We’ve had a couple of people with those kind of products that we’ve spoken to before.
How do you approach finding customers for API’s because they need to be building or integrating at that point. So I can imagine it’s potentially pretty difficult to get the awareness to attract customers, how have you approached that?
Felix Laumann
It’s indeed very difficult. What we have really been been doing is kind of, generally speaking broadly about about NLP and about what can be done with NLP. Then kind of get, let’s call it like business buy in first. Then having okay, like we integrated through an API, because that’s like the easiest way to do it. When you do API first sales, that’s super difficult even to someone who’s coming from business. It’s very much like what value will you bring? Instead of okay, here’s an API and you do visit whatever you want. I don’t think API first really works. It’s probably more of a marketing generating demand.
Matthew Todd
Does that mean in terms of inbound versus outbound that you’re you’re more inbound focused, and getting a lot more customers that way,
Felix Laumann
When we do outbound, but the inbound is brilliant, because if people already know, like, we are potential solution to their problem. It’s not just about like, how we can best serve them and how they probably don’t only contact us, right. So there are obviously a bunch of competitors. Then how we how we can differentiate from them. We try to have as many inbound as possible. We try to speak at conferences.
Quite soon, we will do more marketing campaigns on social media. It’s still heavily outbound reliant where people very often don’t know what that technology can actually do. Then you need to need to explain them, you need to actually show them that they have potential to improve processes.
Matthew Todd
I can imagine, with a new technology or a new technology, going into new markets, enabling new things to be achieved, from a technical perspective, there must be a lot of education required in order to let those people know that actually, there’s something is possible that they didn’t know about before.
Felix Laumann
Yep. 100%. We do regular like webinars, office hours, which are more on the education side.
Matthew Todd
In terms of growth of the platform and features, and advancing it, moving it forward, how do you balance innovating from a feature perspective versus doubling down on the accuracy? Versus adding more languages? I can imagine that’s quite a few different things potentially in competition with each other.
Felix Laumann
We be tried to listen to customers as carefully as we can.
Felix Laumann
That’s very true. We try to listen to customers as carefully as we can. Number of languages, I think we are somewhat settled, right? Which is nearly 100 languages that we have right now. I think we cover like 98% of the world population in terms of mother tongue, or to speak as a mother tongue. Which is okay, it’s fine, there are very few that we actually do not cover and half had customer interest in. We may add, let’s say another 10 or 20 over the next six months. That’s kind of maturing.
The other two are kind of definitely in competition. So improvement on acurrecy is something what you try to do, especially for the speech technology, because that’s where, at the moment, the most, most accuracy improvement can be done. We have a dedicated mini team that only works on that. Then we also have a second user experience team, where we add features, like downloading datasets, like uploading them easily, uploading different formats, and all of that.
Matthew Todd
We’ve seen a lot of AI platforms and language processing platforms emerge in the last couple of years alone. What do you think of the next set of innovations that will be possible building on top of what we’ve now been able to get to?
Felix Laumann
Good question. I wish I knew. Thinking about the general education of the market, right. So people know more of NLP and what can be done with it. I’m thinking that people take like these developments that can be done with the technology more in house. We just need a flexible, flexible platform that can do lots of different things. So I think we did it already in the American and the European market. In the more developing countries that we don’t see yet. Where people really do that heavy technology development in house. But we see it already in Europe.
So people want to have a quite flexible platform, that can do lots of different things. It may be that people have been calling it multimodal learning or machine learning. What means that take more than one source of information in, right? So you can take in from a video, you can take in the images, and you can take a look into sound to make a joint decision, okay? That’s potentially, someone being abusive to other people on the street, right?
So it can take multiple source of information. That at the moment cannot be tackled with one platform, though, companies really need to go into a computer vision company into an NLP company or voice company, and then kind of put these API’s together. I think that’s still quite difficult. So kind of set merging of technologies to one platform, I think that will be will be coming in the next couple of years.
Matthew Todd
I remember reading a few articles and listen to a few podcasts from the founders of Twilio, who provide, you know, SMS text messaging, but also video based communication API’s, I guess you could call them. Similar to yourself, they started off having a platform, and, you know, offering SDKs API’s that people can call to, to use their infrastructure and technology. But the applications that people were building were very diverse, and certainly not the types of applications that they thought people would be building with it and often had never even considered.
It’s like a layering approach, isn’t it? As people start to build things, you suddenly realize what can be done and what new use case is emerging, as you say, combining technologies as well, what else can be done? And it just helps to advance, you know, all kinds of industries by opening up these API’s to people I think. In terms of your journey, so far then with NeuralSpace and thinking of our audience of startup and scaleup founders as well, what are some of the lessons you’ve learned along the way that you think could be useful for our audience to hear?
Felix Laumann
Whoever thinks about starting a company, or whoever started the company, I pay highest respect. So it’s an exciting journey, but it also is, like a very emotional, and stressful journey. Yesterday, I came across a video on YouTube where someone has interviewed who are now billionaires who started to start a companies, when they were young, they are now 60 or 70.
There was one very interesting quote, where somebody was saying, you should like, like what you do, but actually, it’s actually not good when you’ll be absolutely love your company, because it comes very often to a stage where the company owns you, instead of you owning the company. What he meant by that is that you make decisions that just keep kind of the dream of the company alive, what could maybe lead to not tackling the right market or not never achieving product market fit right, because you have to kind of set imagination in in your mind of the world that the world evolves in that in that sort of direction, but actually never never goes in that direction right.
Felix Laumann
Having an objective mind on your ideas and on your hypothesis on how a market is reacting to your product is extremely healthy.
So, when you absolutely like love the company and the company that idea of the company owns you and take control over how you make decisions, you lose that objectivity, which is very healthy as a founder. So, you should own the company that means you should let it move in a direction that you believe is right. That you have assessed the object as objectively as possible. He was saying like, I liked what we do, what we did, but I followed that idea because that’s a huge market value and huge amount of money for him personally and in the end. That was really what was driving it. So people say okay, money should not be the number one reason why you start a company because we need to believe in the bigger and so on. Definitely, you need to believe that the world is moving in the direction of what you are then partly building. But having an objective mind on your ideas and on your hypothesis on how a market is reacting to your product is extremely healthy. That requires a lot of challenging yourself, it can be actually very painful. At least it was for me. So I had that strong belief and then it doesn’t really work out that way. So you need to do a pivot or mini pivot.
But it requires solely questioning your own assumptions on how you believe the market is reacting or how you believe your initial target customers are reacting to such product. Well, they actually react quite badly to our product. So we need to position ourselves differently. Luckily, the technology is quite flexible.
There may be lots of sleepless nights, and it gets very personal. So don’t believe in a work life balance. You will think about your company as a founder, yeah, like nearly 12 hours, probably even more every day. You need to merge kind of your private and your and your professional life, I would say. Most employees are very close friends of mine, for example, which is nice, which makes it dooble,
You take some sacrifices for it, hopefully, you’ll really fulfill what you have been hoping to achieve. Then it’s probably worth it. But yeah, it doesn’t come for free.
Matthew Todd
I think that’s really good for people to hear that. I like what you were saying about the attachment to the company. But I think also to the product that you’re building, as well, I think a lot of founders can start off doing the right thing spoiling a problem in the market, but then immediately jumped to a solution.
Then through confirmation bias and everything else just, you know, seek out information that confirms the solution is right. Sometimes, not always, but sometimes not truly listening to market feedback. Then being slower to respond than is ideal, or potentially limiting opportunity based on what the market is telling them.
Felix Laumann
What I’ve seen a couple of times already, unfortunately, is that people say the customer is just not ready for my product yet. Or I need to educate the customer more so they like my product. It’s not scalable, right? You can’t acquire 1000 customers, and you need to educate every single customer how to use the product, or what is the value of your product. So it’s much quicker, although it’s personally more difficult. Why don’t you adjust to the market, right?
So instead of trying to really educate every single customer, just go into that direction. Maybe you have not a product as fancy and as technologically advanced. But it’s what the customer market needs.
They need time to adjust and be comfortable with a new technology. Because change always is something very uncomfortable. It just requires, especially for larger companies, it require us to have processes slightly rewritten. You can’t then kind of challenge existing processes too much. But you can’t move like what about 10 years ahead and then, like ask everyone to adopt technology.
Matthew Todd
I think patience is so important, I think in attempting to get traction or to build at a scale or whatever goals founders have set themselves, I think people can often try and rush things. We need to do more marketing. We need more sales. We need more inbound, we need more outbound, whatever it is they can sometimes seek those results too early and then get frustrated and maybe changing tactic.
Then when they don’t immediately see the results that they want. But I think sometimes you do need patience and time for the markets to engage with you, to adapt, and for you to take those learnings as well. If you are willing to be patient than actually, you know, even over a course of six months, you can end up in a far better position than if you’re just trying to rush it every week. And every day,
Felix Laumann
I totally agree. There was one example. Basically, what it enabled people to do is kind of writing a paragraph and then there’s a full website generated for you. Right, which is fascinating, okay, you write a paragraph, and you do some kind of language model in the background, understand the topic and then generate images. I doubt any company actually put all of our trust into into such a very logical, abstract technology that you lose any kind of control.
So what works now for web development is probably something like Wix or WordPress. You have a no code tool, but you still have very much control over everything that happens. There’s not a product market fit. Potentially, whatever, in 10 years, people really trust and people just want to generate a website in five minutes. But it’s an example of okay, like, there’s too much innovation happening at once and I don’t think in many markets, many companies actually able to or willing to take that risk.
Matthew Todd
I think some people, especially with the lot of AI that’s emerging at the moment can try and take it a few steps too far. As you say, like automates everything for me, whereas actually, the real value is in, do the boring bit of getting started and get me 70 80% of the way there. Then let the person take over and give them tooling that augments and enhances their experience rather than replaces it, I think.
Thank you for taking the time, I think it’s really good to hear, you know, both sides of things really, from you know, how you’ve been able to build up neural space, the technology, the way that you’ve been able to differentiate from those big technology players, but also then, you know, hear some of those challenges that founders do have to address as well.
I look forward to seeing how NeuralSpace moves forward and maintains that that differentiation against those competitors. Thank you for joining me on this episode of Inside the ScaleUp. Remember, for the show notes and in depth resources from today’s guest. You can find these on the website insidethescaleup.com. You can also leave feedback on today’s episode, as well as suggest guests and companies you’d like to hear from. Thank you for listening.