Sonia Kampshoff

Welcome to Working With Languages. My name is Sonia Kampshoff. I'm your host and a multilingual digital marketing consultant. On this podcast, I talk to people like us who use their language skills in the work they do in the widest possible way. Together with my guests, I will discuss why languages are even more valuable in the age of AI and how we can all make languages a meaningful part of our career journey. Let's dive in.

before becoming a localization manager. I use their free translator app almost daily, and I'm impressed with the quality of their output. And this is only one of their products. In this conversation, Elena tells us about her studies and her career path, how D-PAL has developed over only a few years, and how they merge linguistic quality with AI and data security.

Sonia Kampshoff

Hello, Elena.

Elena Carpanese

Hi there, how you doing?

Sonia Kampshoff

I'm great. It's really good to have you here today. So as you know, I normally like to start the conversation by asking my fun question. What is your favorite word or phrase in a language that you speak?

Elena

So I thought about this and I was really struggling because I just narrowed down to one. I have two. So one for German and one for Spanish. Those are the two languages that I studied at university. The German one is a word that I actually end up using when I speak English or Italian as well, which is so to be gespannt, gespannt sein, it's a word that...

I don't think you can really translate because it expresses both excitement, but also kind of worry at the same time on something that is happening in the future. So sometimes it's just that like sweet spot of being excited, but also a bit doubtful and worried. And so I do end up using it in my everyday life.

And the other one for Spanish that I really love, and I think it kind of resonates with an Italian and the Italian lifestyle is the word, Tardeo, which is a mix of tarde, which is the afternoon evening, and tapeo, which is to go for tapas, basically. And it's this, you know, I guess tradition that they have in Spain to go out for tapas in the evening and then see where the night takes you. So staying out with friends around a table, having drinks and chatting, which I really love about the Spanish culture. And it's very similar to the Italian culture as well.

Sonia Kampshoff

very nice. I must say I was expecting an Italian word.

Elena

I was thinking about that as well, but I feel like I'm too biased. I never really fully think about my own language in that lens. I think I just take a lot of things for granted. So these two, I think, are the ones that kind of always fascinated me in the languages that I studied.

Sonia Kampshoff

Yeah, they're good ones. So you speak a lot of languages, but you originally come from Italy, right?

Elena

Yeah, correct. So I speak Italian, obviously native level English studied throughout school, and then I ended up studying in the UK and I studied German and Spanish. So I kept the language education going. And that's always been the thing that I wanted to study, the thing that I wanted to know about. And still to this day, I'm picking up languages never fully committing to them. So I started with French and then Norwegian and now I sort of want to go back to French but this year I need to pick one and stick with it. So we'll see how that goes.

Sonia Kampshoff

Great. So you studied in the UK, you said. How did you find the experience studying in the UK? And I find it interesting that you studied also Spanish from English, which is probably very different than studying Spanish from Italian in Italy.

Elena

Yeah, totally. I loved studying in the UK. It was such a different experience to studying in Italy. I didn't go to university in Italy, the school experience was definitely very different. I think what I took from the experience in the UK was a lot of independence and forming your own critical thinking around, you know, different subjects and different matters. It was definitely very interesting.

The combination of learning, for example, a language like Spanish from English, we would do a lot of translations from Spanish into English, which for me was very counterintuitive because it's just Italian and Spanish are so much more similar and closer to each other. So the translation would have been much easier from Spanish into Italian, but it really helped me stimulate my brain in a different way. With German, obviously, as you can imagine, it's quite a good transition in a way.

But yeah, it was very interesting and also very interesting to see who studies languages in the UK, being a very, you know, one of the, you know, countries in Europe that is less exposed to other languages. It was still very inspiring to see a lot of people in my course who decided to go down this career path. So yeah, very good experience overall.

Sonia Kampshoff

And while you were at university, I didn't do my bachelor in the UK. Did you have an idea why you were at university, what you wanted to do afterwards in terms of work? Did you want to work with languages? Did they help you?

Elena

had no idea, to be honest. I was never somebody that knew what they wanted to do with their career, and I honestly still don't know. I've always been somebody that took opportunities that presented themselves to me and kind of shaped my career that way. I think I knew I wanted to work with languages to a capacity. was always, that was why I ended up studying languages and that was what motivated my choice.

I find them super interesting, I find culture is very interesting. I always wanted to be able to speak a language so that I could experience the culture and the people through my own lens without having to rely on somebody else's translation. I wish that was true for all the languages in the world. Of course, I have to rely on translations for the majority of the languages, but for a few of them, I can make my own opinion based on experiencing art, literature, and different cultural aspects.

So when it comes to my I never really gave it a big thought, but I did end up very much in a language-based career, which I do love. Yeah, it's still very exciting to me.

Sonia Kampshoff

Brilliant. And you're now working at DeepL. Was it your first job out of university? Was there something in between?

Elena

So I, yes, it's my, let's say officially my first job out of university. I was always very proactive while I was studying and I would do internships during the summer. I also spent my year abroad while I was studying in Germany and I worked in localization then. So I had the opportunity to do a work placement. And so I decided to do one in Cologne in Germany and I worked in localization, which was then a good transition into DeepL because when I was hired coincidentally in Cologne again, I went back.

They were looking for somebody with some localization experience. So that was very, came in very handy for me. And yeah, and since then I've been at DeepL. My, my role has changed. My job has changed throughout the last five years and the company has changed a lot. So it's been very exciting and very full of surprises, I'd say.

Sonia Kampshoff

Before we go into what you do exactly at DeepL, can you give us a very brief overview of what the company does? And I'm aware that it has changed a lot over the course of the last two years.

Elena

Yeah, so DeepL is a language AI company fundamentally. So it's a company that provides a language service primarily to other businesses. So the main model is B2B and it's a language service that's very customizable. So at the basis of it is machine translation. Nowadays, obviously, LLMs and it has grown a lot, as I said, in the last five years, it started off as a Cologne based company when I joined, pretty much everyone was hired in Cologne. And now we have over a thousand employees across the world. going from the U S to Japan across Europe. We have different offices and we've also developed the amount of translation and language products that we offer. And most recently we've embarked into the world of agentic AI, which is very exciting and very, yeah, new for all of us.

Sonia Kampshoff

I can imagine. Going back to the beginning, can you tell us what you worked on at the beginning? It was only five years ago, but in the world of AI, that's a very long time ago.

Elena

Yeah, indeed. So when I started my role was language and localization specialist. So I was working on two main tasks, I would say two main parts of my role. One was localization. So something that most people are probably used to. I was translating and localizing all of the content that was produced by the company from English into Italian.

So that means all of the UI, anything that you see on the website, any marketing content that would go out to the public or to our customers needed to be localized in different languages. So I was doing that. And then the other side of my job was working with the research team. So there's a team of researchers obviously working on the language AI itself, and they needed to have someone in-house specifically for Italian to be a sort of consultant when it came to the quality of the translation, the quality of the data that we were feeding into the neural networks.

And then when LLMs came, there's a lot of testing that goes into comparing the different services and translators out there. So that was more of a kind of assessing and data-based role, which was equally exciting. But then at some point, because the company was growing, I had to sort of decide which way to go.

And I stuck with localization, which has been my sort of first passion and my role developed even more. became first team lead and then manager in the team. And so now we have a team of 12 of us in total covering different roles. Some of them are translators and still do localization. Some of them have more operational roles, but what we do basically is the same. So we...

We work with different stakeholders within the company to localize whatever content they produce. So that could be, as I said, anything that you see on the DeepL website, on the DeepL app, anything that goes out via our marketing channels is localized into many different languages. And we also, most recently, which is the thing that's most exciting for us, we are acting in a way as consultants to our product development teams to make sure that the products that we create and put out there are very much in line with the language service industry.

So it's a little bit different to what translators and localization specialists are used to. But I think, as you said, in the age of AI, I definitely see this as a new way for us to evolve in our roles and to change the way that we approach vocalization.

Sonia Kampshoff

So how many tools do you actually have? So I normally use the translator, but I know that you have so many more tools.

Elena

Yeah, I'd say we, I'd group it into like four main pillars. So there's the whole, what we call language AI. So what you've mentioned, the translator is the main product there, which could be accessible through the website, but you also have different desktop apps. We have web extensions, we have mobile apps even. And that's sort of the, the main, the main product.

It also has customization tools, so you could use glossaries, personalization tools, you could have different profiles depending on what you use it for. The use case depends very much on the company and what they use Devel for. And the other three are probably not as known to the single user, which would be our voice product, which is voice translation.

So that's also quite exciting and we're constantly working on it. It's speech to text, but also speech to speech. So there's a lot in that regard that's kind of still developing. One is also API. So if you are specific, there's a lot of companies that require that service from us. So the possibility of basically integrating our machine translation into their own products and tools via our API. And then as I mentioned earlier, the final one that is the newest one is our agent. So I would say four main pillars of the products. And they're all interconnected. Obviously, the technology behind it is the same.

Sonia Kampshoff

Yeah. And I'm assuming that the sequence you mentioned them is also the sequence you brought them out, you launched them. So probably first the text translation, then voice and then agents.

Elena

Sonia Kampshoff

Who are your clients?

Elena

Hmm. So many different profiles. So I would say the majority, as I mentioned, is businesses and it's businesses that have multilingual needs, right? So it could be that they work in different markets and so they need to cater to the different markets. I'm thinking, it could literally go from e-commerce to big manufacturing companies, legal companies, healthcare, life sciences, other tech companies.

There's also companies that have requirements to have internal communication translated. So they have internal documents that need to be translated because maybe they work across different countries or even multilingual communication across colleagues that's sometimes really needed, right? In meetings or in chats, emails. So that's something that customers really use very often. And then I'd say, yeah, I'd say these are the main sort of use cases that I can think of.

There's also, you know, countries where, you know, multilingual communication is a must, is a legal requirement, I'm thinking of countries like Canada or Switzerland. So a lot of companies in those countries have these kind of even legal requirements. And one of the and strongest points, selling points of DeepL is data security, which I think nowadays is such a big topic. And so for companies, this is generally really, really important. So yeah.

Sonia Kampshoff

Of course, And it sounds like once you acquire a client, it's a really long-term relationship that you build with them and they continue using your tools.

Elena

Yeah, that's, that's, think what I've seen throughout the five years. And I think that's generally really, I would say reflected also in the company culture. I mean, I've been there for five years and I also know many, many people, colleagues that started around the same time. Some people even before me that are still here today.

So I think it's a big indicator of a company that looks after their employees and their customers in a really good way. So that's the impression that I have. My role is not customer facing, so I don't really speak to customers firsthand. But of course, we share success stories in the company and it's really good for us to have an idea of what's going on outside of the company. And so that's the experience that I've had.

Sonia Kampshoff

Very good. When it comes to the company itself, can you give us an overview of how large it is, which departments there are? And I'm assuming that all the R &D and technology side is very heavily present in the company.

Elena

Yeah. So it's really difficult to map it out fully because as I said, it's over a thousand employees. It's the general sort of department that you would have at a tech company. As you said, R &D very, we're very big on that. Of course you have to be when you work in AI and tech, it's such a competitive industry and such a fast based industry where you have to really invest on creating technology and products that will make you be able to compete with all the options that are out there.

So my team currently sits within the marketing department, but we also have departments like product, engineering, research. As I said, we have the customer department. They deal with everything customer facing sales. So anything that you would think of in a standard tech scale up, that's the sort of landscape at DeepL as well.

Sonia Kampshoff

That's great. And what's the gender split?

Elena

Ooh, that's very interesting. I would have to look it up for you. I want to say it's a fair split. I think the company, we have a lot of ERGs in the company that really put a lot of effort in making sure that there's a lot of inclusivity. So whether that's, you know, gender neurodivergency, ethnic differences. So I think that we are doing quite a good job as employees to kind of engage with this and also make sure that leadership is keeping this in mind, which they are.

So when we hire, for example, this is something that we keep in mind. We have infrastructures within our hiring process that make sure that we hire fairly. And I think this is something that was pushed a lot both by leadership and the employees in general. think there's a general sense of we always need to do better. There's always something that you can do better in this regard. When it comes to actual numbers, I am not sure I would have to, I have to look that up for you.

Sonia Kampshoff

It sounds anyway that it's fairly balanced and that you're happy about it. It's not a red flag, so to say. So you worked in the Cologne office and now you're in the London office. Where are the other offices located worldwide?

Elena

Definitely

So we have offices in the US, there's London in the UK, a few ones in Germany, and then we have an office in Tokyo, Japan. I hope I'm not forgetting anything, but I think that's pretty much it. So we work across different time zones, which is really interesting. That's also something that we have to keep in mind when having meetings.

But I think for me personally, it's what makes it really exciting, especially when we meet as a company or we have you know, off sites where you meet people in person. We work primarily remotely, but going to the office does make a difference and meeting people from other countries is always very exciting and keeps it very, as we said, multicultural and very interesting. And as a company that wants to break down language and cultural barriers, I think that's super important to maintain this diversity.

Sonia Kampshoff

So you mentioned that you have a team, a small team of linguists that you work with. Do you have a lot of other linguists in other departments maybe? And you also work with external linguists as consultants.

Elena

We do. So my team specifically, we, as I said, we have a small team of in-house localization specialists, but we cover a lot of different languages and we work with external freelancers. So we have relationships with different freelancers. We have a network that we manage internally. There's also other teams. So as I mentioned earlier, when the job that I was doing at the very beginning, where I was working closely with the research team.

That team still exists. It's just a separate team now and they work in a different department. But they still work with the research team on the data and on the translation quality. And they also work with freelancers. Their network of freelancers is much bigger. And they basically rely on them for all of the tasks that I explained earlier and even more things. So they work on projects to always ensure that whether we're adding new languages or we're improving already existing ones in the platform, that they are basically providing, we're providing the best translation quality. So yeah, it's a really big network. I'm not sure how many people, since it's not my direct team, but yeah, we do rely on a lot of freelancers and linguists and translators.

Sonia Kampshoff

How many languages do you support at the moment?

Elena

I think the translator offers more than a hundred. I would have to double check. We recently launched a really big number of translations all at once, which was very exciting. It unblocked a lot of different opportunities. When it comes to localization, I believe we localize in around 13 languages. Five of them have in-house localization specialists. So yeah, I'd say again, I would have to double check numbers but I'd say the translator office is around just yeah over a hundred.

Sonia Kampshoff

How do you decide when to add a new language and which one to add?

Elena

So that really depends. It really depends on the requirement from customers. That's always a really big driver, talking to customers, understanding what they need. But also, sometimes it really depends on what data is available, how good the data is. We always want to make sure that the translation, the output, is the best that it can be. That's really, both factors, I'd say, are equally important.

But I think nowadays with, you know, LLMs and all the different resources available out there, it's becoming easier and easier to unblock languages. The downside is that the quality needs to be checked more because there's, as you know, more available out there and not everything is good. So yeah, it really depends.

Sonia Kampshoff

Can you tell us a little bit more about the agent AI that you are developing?

Elena

That's actually really exciting. My team was one of the first ones to test different use cases when the agentic team was still working on it. It's really hard to explain what it is because what we say is that it can do anything. And when you say that to people, it's really hard to visualize what it can actually do because what do you mean it can do anything?

So one of the things that we want to focus on with my team, especially this year, is to figure out what are the best use cases to sort of help the public understand a little bit what it can actually do well. When we launched the agent, we also launched a website which is called Default AI Labs. And there, we've explained and actually shown in practice what the agent does. Yeah, so it's still...

How do I explain it? It's very versatile. And to be honest, it doesn't only apply to the localization industry. I think there's even other industries that have been automated way less in the past that can benefit from it even more. I think localization is one of the industries where we've had to deal with automation starting from machine translation, but also, know using TMSs and workflow orchestration, all of these different tools that now exist, where a lot of our workflows are already automated.

But there's other industries, like I'm thinking, for example, the legal industry, that have always kept it a little bit more old school. And I think they are the ones that could benefit from the agent even more, because effectively what it does is it automates and does tasks for you that are very repetitive, you wouldn't, that take time away from your actual work. So while you, while the agent does those, I'd say boring, repetitive tasks for you, you can actually work on your real work and focus on your real job.

Sonia Kampshoff

So what is the intersection then between agent and language? Can you give us some examples?

Elena

Yeah. So we were, for example, testing, first of all, as I said, how to automate some of our workflows. But then something I looked into, which I thought was really interesting was how can we audit our translation output using the agent? Can we use the agent to have a look at, for example, our interface in Italian of the DeepL website and have the agent assess whether the translation is effective, comparing it to the English, whether the terminology used is easily understood by the user, whether the user journey is intuitive from a linguistic perspective.

So this is something that would take you a really long time because you would have to really

go through different parts of the website, follow the user journey yourself. But having an agent do that for you saves you so much time. And also, it gives you an evaluation at the end, which is really interesting and something that maybe you wouldn't have thought of. So that's just one small use case that we worked on.

Another one that, for example, we show in AI labs is terminology gathering. So you could ask an agent to, for example, go onto a database, go through a bunch of different, let's say legal articles, collect the terms that are used the most, and then compile a glossary for you, which you can then use on your legal translations.

So something that would have taken you hours because you have to read through many articles, figure out how many times a word is used, can be easily done by technology. And at the end, you just have something that's ready for you. You can think of it as sort of an advanced version of generative AI. So instead of just producing something for you, it can navigate the web for you and actually do things for you. So check out different pages, select different things and yeah.

Sonia Kampshoff

Interesting, very interesting. And so you mentioned the legal industry, but it kind of applies to all industries, doesn't it?

Elena

Yeah, that was just an example. It can really be applied to anything. And I think it's really hard at the moment, not just for us, but just for people in general. There's a lot of different products that are coming out of AI companies. And sometimes we don't know necessarily how to use them best. And I think right now, this is the time to really test the limits, see what they can do well, what they can't do so well.

And I think the use cases will differ depending on what you do in your everyday work. So it's really hard to give a how-to guide. Sometimes we don't even know if they can do something until you see it done and it really surprises you. Wow, okay, this was possible. And maybe things that you didn't think of.

And I think sometimes, the technology, you can guide that technology, but sometimes that technology guides you in a way. So yeah, this is the sort of approach that I have when I use AI myself in my job to kind of approach it with an open mind of let's see what it can do.

Sonia Kampshoff

Elena

So the idea is to continue working on these four main products. I think there's going to be more developments on all fronts, which I'm very excited about. As I mentioned, more language AI products. I think continuing down that path of really establishing ourselves even more as a player in the language industry. And yeah, exploring agentic, I think that's probably something that the majority of AI companies are doing at the moment.

Sonia Kampshoff

But the focus is still to include language in everything that you do,

Elena

Absolutely, yeah. I think that's always been the core of the company and it will continue to be the core of the company.

Sonia Kampshoff

Is there anything else that you would like to add that I haven't asked?

Elena

No, I think we've covered a lot of interesting things. So thank you so much for having me. It was really interesting to have this conversation and also sometimes hear what questions people have. You might think that people are interested in something and then it turns out that there's something completely different than they're interested in. So it was great to have the conversation with you and yeah, thanks so much for having me.

Sonia Kampshoff

It was a pleasure. was a really interesting conversation. Thank you so much for coming.

Elena

Thank you.

Sonia Kampshoff

In only a few years, DeepL grew from a German startup to a global communication service provider with a solid standing, all the while keeping languages and communication at the core of what they do together with AI. I'm looking forward to seeing where they go next and how they will further develop voice translation and agentic AI. If you enjoyed this episode, please give it a five-star rating on Spotify, Apple Podcasts, or any other platform where you listen to your podcasts. And if you have a comment, simply leave it below.