I dove into the challenges of running a highly curated global nutrition data API with Sebastian Loch, Business Development at FatSecret. He talked about his experience of running a modern API operations that has been around for years, but he also shared his challenges in putting artificial intelligence to work as part of their platform, as well as the uncertainty of providing high quality data via API in a world where AI consumes everything.
API Evangelist Conversation with Sebastian Loch, Business Development at FatSecret
Transcript
All right, this is Kin Lane here. I am once again with API Evangelist Conversations, where I just spend a little bit of time each day having conversations with interesting folks doing interesting things with APIs. Let’s see who we got with us today. Hey there, who are you?
Hi Kin, my name is Sebastian.
What do you do, Sebastian?
I’m the head of business development here at FatSecret, and I look after our platform API business.
Platform API. So I’m going to start with a simple question, a pretty basic one. Why do APIs matter to your business?
Well, I think APIs are the backbone of modern-day IT infrastructure, and to build scalable IT infrastructure and solutions, especially when you touch so many lives like we do in the health space and the nutrition space and weight loss space, I think you need to have robust APIs in order to power these solutions both on the web and for mobile apps for millions and millions of users every day.
Agreed. They’re pretty critical to everyone’s business, but sometimes people aren’t always aware of it, and they’re not easy either. So when it comes to running APIs, what’s your biggest challenge right now?
We operate in this very niche market of nutrition data that most people are not familiar with, and our API is embedded in so many modern-day health and fitness applications, medical applications, that no one knows about. It’s fully white-labeled. So it’s this niche market, and it’s really difficult to make a proper business out of it, and it took us probably 10 years, really, to be frank, to build a proper API licensing business, and at the same time doing it at scale globally. So we operate across 56 countries and we provide verified nutrition data. So scaling this business globally and finding the right partners and the right customers is probably the biggest challenge.
So I want to get into the technical of how you do things, and you mention global operations, but I kind of want to fast forward and talk about, I think for me one of the biggest concerns, opportunities, it depends on how you look at it. You mentioned you’ve done all this work putting this data together. What does it mean with an AI or artificial intelligence landscape that we kind of find ourselves in today? What does it mean?
It’s a very good question. Potentially, short term, this is a big threat. So we have been curating, verifying, and publishing nutrition data for the past 17 years. We started back in ‘07, and back in ‘07 there was no nutrition API out there, you couldn’t just tap into an open API and plug it into your mobile app or your web app and receive verified nutrition data. And that led us to doing this ourselves. So we built systems and tools and processes to curate nutrition data, and then publishing it on our website, which then had many companies, corporations come up and approach us and ask if we can make this data available to them. So we started very early on with APIs, and fast forwarding, as you said, to the modern-day age of AI, there is now a potential threat to our API business, whereby developers will not use us as a verified source for nutrition, but they will just ping OpenAI or any other LLM, really. So short term it might be a threat, but at the same time we know that a lot of LLMs have their downfalls, they hallucinate, they often don’t have good data when it comes to international markets. So for example, if you want to know how much sugar is in a Snickers bar, you can probably ask ChatGPT and it might have a decent result, but if you want to ask the same question for a product in Brazil, in Indonesia, in Japan, in Korea, in Germany, in Australia, the results are not that good. So there is still a need, we believe, for proprietary data to be providing you with verified nutrition information. But at the same time we know that AI needs training data, and I think that’s the second part of the conversation, whereby we know AI is already struggling to be provided with reliable and really verified training data, and I think that’s potentially another avenue that we might embark on in the coming years.
So you have a pretty sophisticated API portal. I’ve been through your site, you’ve got a lot of all the common building blocks I would expect from an API program, and you’ve clearly been serving web, mobile, kind of diverse applications. But what would you have to do to kind of prepare for this? I mean, you talked about business development, finding new customers. It seems like you still got to have a public API and have this available for people to test drive and see, but you’re going to need some more solid contracts with people to be able to build these relationships, right?
Yep, 100%, and that’s why we’re trying to innovate and build on top of our existing API substantially. So we’ve probably neglected that side of our business for maybe six, seven, eight years, and we focused possibly too much on our consumer side of the business, where we run a calorie counting app, it’s called FatSecret Calorie Counter, by FatSecret, has about 130 million downloads. But now we’re trying to leverage, for example, AI capabilities to build on top of our existing API platform. For example, we want to embed, or we will be embedding, image recognition very soon, natural language processing, and many other features that were kind of out there in the past, say five, six, 10 years ago. You could use large language models, weren’t really the thing back then, but machine learning obviously did exist 10 years ago already, so you could build somewhat very basic natural language processing capabilities and very basic image recognition technology. But now obviously in the modern-day age and the modern AI world, it’s much more easier to build really solid and robust features that help people look up nutrition information, log their foods way more easily, to reduce friction. And I think that’s where the future lies. So we really have to push ourselves every day and innovate harder and faster than what we’ve done in the past to stay relevant.
Yeah, so it’s a whole mix. I mean, the API relationship to AI is, AI is trained on and uses API, and then you have to be able to use those language models, you have an API that people would access it. But you’re also talking kind of a third paradigm, is where you’ve got to build other models and use other AI services and kind of layer them into what you’re doing to kind of augment and enrich and make your core services more valuable.
Yeah, 100%. And we tried to do it ourselves, especially in the day, say in 2016 we started our own image recognition technology, so we hired our own team to build machine learning capabilities, and that was sort of pre-AI, or what people now perceive as AI obviously was to some extent that. But what I’m trying to say is, all these things that people have built in say 2012, 2015, 2016, 2018, even 2020, is largely irrelevant now. The modern-day LLMs have become so good that most things that you build in house, whether you have two people, 10 people, or 20 people working on it, we know that the modern-day big tech giants, they have 100 people, 200 people sitting there doing this with much greater scale, much greater impact. So there’s so many really robust out-of-the-box solutions now that you can tap into that will make things for the consumer, looking at the consumer, it’ll make life for the consumers so much easier. And we try and always look back, what’s the best possible solution for the user out there.
Yeah, I think I’m telling all of my clients and would-be clients that you just got to really have your API act together to kind of play in this world, both as a producer and a consumer. Because your raw resources have to be top quality and as curated and well done as what you guys have done with your food database, and then you have to be able to consume and be efficient at doing that. So I mean, how do you keep up, moving to the tech side of things? Can you share anything on how you run your APIs, or how you do APIs, that would help people understand how you guys keep up?
Yeah, I’m not technical myself, so I’m not a developer, but obviously I work with developers almost 24/7. So we built a pretty standard REST API in 2010, it was a free and open API that people could tap into, it authenticates via OAuth 1.0 and 2.0, and we host it with AWS, so very, very low latency, and it’s got great coverage all over the world. We have, as I said, a global business, so we have clients, customers, API developers almost in every country of the world, close to 40,000 developers now using the API. So it has to be accessible from anywhere with minimal latency. And then we have, in terms of the API calls, we have a lookup which we call food search, and we have barcode scanning, we have autocomplete search, similar to what Google does, where you can autocomplete a query, like when a user, say an English-speaking user, types in “bana,” we know they’re looking for a banana. And then that will be made even easier with natural language processing and image recognition. But yeah, high level, it’s a very standard REST API, hosted with AWS, that provides all sorts of standard authentication, and then we have what we call three-legged OAuth as well. It’s a protocol whereby a lot of health and fitness apps in particular use three-legged OAuth to sync user data back and forth, like Apple, you know, Apple Health, Google Fit does it as well, whereby a user might use a mobile application and they log their foods or they keep track of their weight or keep track of their exercise, and they can sync that data back into a public repository like Apple Health and Google Fit, or they sync it straight back into another third party app, which is something we provide and power as well.
A whole lot in what you just said, standard kind of modern API, really getting interoperability, which I think is important in several industries you work in, healthcare, multiple. And then you started by saying, oh, I’m not technical, but then you rattle through some pretty technical things. So one of the things I’m going to focus on moving forward with my next kind of round of storytelling is, I’m really trying to reach product managers and business stakeholders, try to convince them, and outside of the Silicon Valley kind of tech echo chamber, that they need to pay attention to APIs. Why do you, so what got you, learning all those technical details and understanding it the way that you do?
Yeah, I think you have to, I think you simply have to over time. I know it takes time, especially when you don’t have a developer background and you’ve never completed an IT degree. And I’m very fortunate to have good colleagues who are very technical and can explain things and can help me understand things. Like for example, I think I saw that you’ve been active around Postman as well, obviously you’re very active on GitHub, and then I think you have some sort of background with running Postman libraries as well. And I embarked on this Postman journey maybe two years ago when our CTO helped me set it up, and yeah, it’s been a lifesaver. I can, without being a developer, without being technical, I can use Postman, I can use our API, and I use any other APIs for testing and evaluation purposes via Postman. So that’s amazing.
Yeah, it’s kind of a browser for APIs, and it makes it a little bit easier to kind of navigate this whole world, I agree. One last kind of more personal question, I guess. Why do you do all this? What keeps you coming into work every day?
That is a good question. So the majority of our work that we do is in nutrition and weight loss, and when you look at the global landscape of what’s happened in the last few years, like the global obesity rate has tripled since 1975, and 60% of Americans are now overweight or obese. The obesity pandemic is on the rise, and the UN calls it a pandemic, actually, it’s been labeled as a pandemic. So the problem we’re facing as a society and as humanity as a whole is unbelievable, and many people haven’t even realized how much impact food and nutrition, daily nutrition, has on your overall health and well-being, and the rise of chronic diseases not just in the US but globally is out of control. So we really try and build effective weight loss tools. That’s how we started, I mean we started back in ‘07 as a social network for dieting, that’s what we called it back in ‘07, when there was the days of Facebook, when Facebook started and became big, so a long time ago, but we’re still here, and we try and build effective weight loss tools, because we know it’s a growing global problem. And again, the rise of obesity and the rise of chronic diseases, largely nutrition-related, unfortunately, are almost out of control. So we’re really trying to have an impact, and we’re touching the lives of many users already, with our API has just over 700 million API calls every month, and we have, as I mentioned, more than 130 million downloads for our own app. So there’s a large global audience that we serve that we try and help, and at the end of the day, that’s our motivation.
I like it. I love good quality APIs with good quality data doing important, good things, with people who care. It’s like exactly what I love hearing more about, because I’ve been doing this full-time for 15 years, and you see a lot of APIs that aren’t quite useful. So it pleases me to hear one that’s having such a great impact. So I appreciate you coming by today and sharing your story.
Thank you so much, thanks Kin, for having me, that was great.
All right, appreciate it. All right, ah, I do love a good API that has good value and is trying to make their way in this world. I think the AI conversation is something that’s super interesting, and trying to understand how people are going to make their way in this, as the next generation of apps is going to be pretty critical. So, well, thank you everybody, I appreciate y’all tuning in today, and we’ll see you next time.
