“I was among the very first people Facebook hired internationally.”
EPFL alumnus Rodrigo Schmidt has a unique insight into Facebook and Instagram: over the last 15 years, he’s directed engineering at both.
How were you able to join Facebook as a young EPFL graduate?
I’m originally from Brazil and went to EPFL for my PhD in Computer Science. I graduated in early 2008 and spent a few more months at EPFL conducting postdoc research through an innovation grant. I really wanted to stay in Switzerland, but my options were limited and the opportunities there were not the same as in bigger high-tech clusters. So I interviewed with Microsoft, Intel, Yahoo! Labs in Barcelona and a few others. And then there was this startup, Facebook, with only a few hundred employees. As a funny story, at the time, they were not hiring internationally so they wouldn’t pay for my plane ticket to California for an interview. But it so happened that I was already going to the US for a conference in Seattle, so they said: “Ok, we’ll pay for your ticket from Seattle to Palo Alto.” I got the job offer on my trip back to Switzerland, and officially started at Facebook in October 2008. A friend of mine who also graduated from EPFL, Alok Menghrajani, and I were among the very first people Facebook hired internationally.
Now you’re a Senior Director of Engineering at Instagram. What brought you to Instagram and what is your role there?
I was hired by Instagram twice actually. The first time was in 2013, a year after the company was acquired by Facebook. I was a happy Instagram user myself and wanted to join the company, so I offered to help. At the time Instagram was growing tremendously and its engineering systems still had a lot of room to grow. For example, the content recommendations made to people were not personalized, and some simplistic solutions based only on the number of likes allowed the proliferation of low-quality content, like memes. Even the fact that the user feed was time-based made it too easy for some accounts to game it – you just had to keep posting content all the time to reach the top of your audience’s feed. I helped create our “data and personalization” team to improve our product with better data and machine learning systems. As a result, Instagram now offers a richer, more personalized experience. The culmination of that vision were the launches of ranked feed and Stories in 2016. It felt like the end of a chapter for me, and I decided to leave the company. But after a brief stint with another startup, I returned to Facebook. I came back to Instagram in 2018 to oversee our engineering efforts across all of monetization, shopping, content discovery, and tools for creators. The overall vision of our product group, called Interests, is to help people connect with things they love. We are responsible for many products like Ads, Shop Tab, Creator Tools, IGTV, Search, Explore, to name a few.
Of Instagram’s one billion active accounts, 200 million are selling products. Is Instagram becoming more of a marketing network than a social network?
I believe Instagram is a balanced blend! A huge part of why people engage with Instagram is still very social – take Stories, for instance. The main thing most people come to Instagram for is still friends and family. But the whole idea of Instagram is that you can connect with your interests in the broadest sense – so family and friends of course, but also your interests in cooking or sports, for example. That’s probably what sets Instagram apart: it’s above all an “interests” network.
And because it’s focused on interests, there is more space for commercial content in general, even outside of paid ads, by creators, interest accounts, or regular people. A couple of years ago, my wife and I renovated our home. We saw furniture we liked on the interior design accounts we followed on Instagram, but we had to buy the furniture elsewhere since back then you couldn’t shop on Instagram. It makes sense that today you can tap on a picture, select the product you want and buy it directly. We basically streamlined something that already existed.
Instagram is apart: it’s above all an “interests” network.
How will Instagram be able to keep up that success as new competitors emerge?
I genuinely believe Instagram still has a lot of scope to grow. Smartphones will only get better, just as technology like augmented reality is improving by the hour and will definitely play a key role in the way we connect with each other. Shopping on Instagram is just in its infancy. We’re investing a lot in the personal experience people have in the platform. We can still improve Instagram’s features for helping people create content. To give you an example, I have many friends and family members who are active on Instagram but don’t share much content themselves. Stories went a long way, but we can still find new ways to make content creation as easy as we made content consumption. New features such as Reels, which was released in 2020 and lets users record 15-second multi-clip videos, is a step in that direction with content that’s lighthearted and fun.
Why do you think Europe in general, and Switzerland in particular, has not been able to produce a company like Facebook or Instagram?
I don’t think you can point to a single reason. For a business to succeed, you need all the various elements: education, money, talent, investors and opportunities. There is obviously money in Switzerland, and the educational system is top-notch. Both EPFL and ETH Zurich are excellent and I would definitely want my kids to study in Switzerland! But there are fewer companies and opportunities, so retaining talent can be hard. In my case, I wanted to stay in Switzerland and so did my wife, but we couldn’t find the opportunities we were looking for. Silicon Valley has it all, and its ecosystem is now self-sustained, which is extremely hard to replicate.
Augmented reality is improving by the hour and will definitely play a key role in the way we connect with each other.
One of the most challenging aspects of social networks is moderating content and monitoring its relevance.
Yes, this is a major focus area across our entire company and we take integrity extremely seriously. We review suspicious content and remove posts that violate our policies. But since there are millions of Facebook and Instagram users around the world, we need a system that is scalable – we obviously can’t check every single post manually. Where computer engineering, and more specifically machine learning and AI, can help is in making the process automatic by building in classifiers and models that spot malicious content automatically, be it fake news, pornography, violence – the list is long. The catch is that the smarter our technology becomes, the smarter the malicious-content producers also become, because they quickly learn how to engineer work-arounds. So you have to outsmart them every step of the way. What makes me optimistic is that, thanks to how machine learning and technology are evolving, our models are becoming extremely robust.
Let’s take the example of deepfakes, since that’s the feature topic of this issue. What technology are you using to combat them?
Just to be clear, not all deepfake content is necessarily bad. Some creative media can rely on deepfake technology to create an interesting effect without misleading the audience. But if some content was manipulated through deepfake to spread misinformation, there needs to be a way to identify it’s false. The most promising technology to help achieve that, in my opinion, is machine learning. It’s possible to build classifiers that identify fake videos based on huge collections of real and fake media. The problem is that if given enough time, deepfakes could end up being as good as real videos. But for now that’s only theoretical; in practice there are always small things that signal a video is fake. You don’t always know what those signals are, but good models will notice there’s something wrong. And they do so in an extremely efficient way, even if they can’t pinpoint what exactly that something wrong is. Another technology that helps now and can help even more in the future is certificates. Even if deepfakes become very hard to be identified, digital certificates could be used to validate which videos are genuinely real.
Another key issue is personal data protection, and it’s one users are paying more and more attention to. How has that affected your job as an Instagram engineer?
Both Facebook and Instagram deal with highly personal data, so it affects my job significantly. We address data protection at every level, from the most basic – like what servers we use and where they’re located – to the product we offer; that is, the settings and tools that let users manage the confidentiality of their data and content. And we have several security checks at every level. That makes our work more complicated and meticulous, which is a must given the responsibility we have.
How would you like to see social networks evolve over the next ten years?
I tend to be very utilitarian in my preferences, so I love messaging tools like WhatsApp and Messenger. Such communication systems can be extremely powerful in a world where more people are working remotely. I would like to see social networks extend their utility and become places where you don’t just consume content, but also get things done. I mentioned being able to shop now on Instagram for example and social networks could also become platforms for education, where users expand their knowledge. In terms of technology, features like augmented reality – as seen in the trend for AR glasses, for example – and even virtual reality will certainly play a key role in creating more engaging content.
Engineers and product developers are given a great deal of latitude, so it’s essential that they understand their ethical responsibility
As we discussed earlier, today there is greater awareness of the ethical responsibility that comes with being an engineer. How can we better prepare the next generation for this?
I’ve been too far removed from academia over the past ten years to know how things are taught today, but ethics was definitely not a topic covered enough when I was a student. Engineers and product developers are given a great deal of latitude, so it’s essential that they understand their ethical responsibility – and education has a key role to play here. It would help to have more courses on social responsibility, product thinking and product empathy, for instance. And it’s not just at universities: companies need to be accountable as well. At Facebook, we have an extensive onboarding program and employees undergo training throughout their careers on privacy, integrity, social responsibility and more. The goal is to make sure people really own the product they’re working on and understand the responsibility that comes with it.
You’ve been in the Bay Area for almost 15 years now. How has the region and its tech industry changed over the years?
One change I saw during my time here was that the center of gravity has shifted a little from the South Bay to San Francisco. When I first came here, San Francisco was not as popular as it is today with the tech community, which concentrated further south. Around 2010, many tech companies started locating their headquarters in the city rather than in the Silicon Valley. Of course, the South Bay is still the main tech hub with the headquarters of Facebook, Apple, Google, and others. But we have seen a kind of shift. The influx of tech companies into San Francisco has brought younger people and some life back to the city, which is very enjoyable.
Do you still have ties with the EPFL community? If so, which kind?
I’m still in touch with many friends from EPFL and with my PhD supervisor, Willy Zwaenepoel, the former dean of the School of Computer and Communication Sciences. Here in San Francisco, I’m in contact with swissnex. But my connection with EPFL has faded a little over time, so it’s great to be in touch again now. The last time I was in Switzerland was nine years ago, and I’ve been wanting to go back ever since. I can’t wait!
Since this interview, Rodrigo Schmidt has been named senior director of engineering at Facebook AI.