Digitizing and automating identity verification is incredibly complex. Features like facial recognition and ID verification use advanced mathematical algorithms and must be both accurate and fast. Jovan Jovanovic is intimately familiar with all things tech at Incode. With a background in math and computer science, Jovan leads the engineering teams responsible for the magic happening under the hood of Incode Omni Platform. Today, Jovan takes us on a journey from the past to the future as he describes the history of the company and where Incode is headed.
Can you describe your role at Incode?
I’m responsible for several engineering groups including backend, web engineering, infrastructure, and cloud services (AWS, Microsoft Azure, and Google Cloud). I’ll be moving to headquarters at San Francisco soon and will be building a new team focused on research and innovation. That new group will be focused on solving new problems and using technology to push the realms of possibilities.
How did your previous startup experience help with what you’re doing at Incode now?
Prior to Incode, I founded a startup that looked for optimal flight routes given a number of cities and days to spend at each location. Although I didn’t continue with that startup, I gained invaluable engineering experience.
Even when I began working with Ricardo Amper , Incode’s founder and CEO, our subject area was different. We initially focused on facial recognition for photo sharing on social media. Essentially, we wanted to be able to suggest friends to share photos with based on existing photos of them in the user profile. The problem at the time was accurate facial recognition – you needed to have five to six photos of a person, ideally at different angles, in order for facial recognition to be accurate. This was very rare in social media profile pictures.
However, through our social media app project, we gained experience on many things that are relevant to Incode’s current business. For example, we learned how to deal with pain points concerning privacy and a user’s desire to know what images are being shared, or how they can limit sharing. We also developed experience building mobile apps that had to run on devices with poor CPU and RAM. Through that experience, we were able to create high quality facial recognition that was fast, accurate, and worked well on mobile devices. All this became essential knowledge for Incode’s identity verification business.
What were some of the lessons you learned? What advice would you give to other entrepreneurs?
Your technology can be great, but you can’t ignore the business side if you want your startup to succeed. Finding the right market for your product is essential. Timing also plays a part.
With our social media app project, everyone agreed the technology was great. We had huge traction and a large number of downloads. However, retention was low since the market was not quite ready for the concept of photo sharing at the time. But despite these setbacks, we were still able to leverage our experience in facial recognition and reposition the business.
Can you tell us how Incode, as we know it, got started? How did you pivot from social media to identity verification?
Our turning point came when we applied for a RFP (request for proposal) from Citibanamex, the second-largest bank in Mexico. Citibanamex was looking for a vendor for biometric validation of identity that included four components: 1) facial recognition, 2) facial liveness, 3) ID validation, and 4) video conferencing.
From our previous experience with our social media app, we already had expertise in facial recognition. However, now we had to grapple with new challenges like people faking IDs, using OCR (optical character recognition) to read IDs, and validating facial liveness checks. But we got to work fast.
After a short amount of time, we were able to produce a MVP (minimum viable product) with all four components Citibanamex wanted. When we presented our solution to Citibanamex, they were blown away by the speed in which we were able to produce results according to their requirements. They began working closely with us, giving us insights and helping to improve the user experience. They were instrumental in helping with testing as well, particularly around ID validation.
In the end, we built a solution that integrated facial recognition and ID validation during a conference call. During the demo, a user was asked to show their ID during a video conference. We were able to compare that data in real-time with the data captured during the initial onboarding, making an accurate verification without disrupting user experience.
All that technology was new to the market at the time. We still had a few bugs we needed to iron out, but Citibanamex recognized our potential to move quickly and build to client specs. After nine months of evaluations, they officially said we won the RFP!
That was the beginning for us.
What a great story about the foundation of the company!
Let’s move on to some questions about Incode’s current offerings. We get questions often around customer privacy. Does Incode keep the images of people when onboarding?
First, I want to clarify that we do not use or store photos and images of people. What we use are biometric templates . Biometric templates are the output of our facial recognition neural network. It’s a mathematical representation of someone’s face and is essentially an encrypted string of characters under half a kilobyte in size. So we only keep the facial fingerprint, not your actual facial image. If anyone steals the facial fingerprint (which is encrypted anyway), they cannot recreate your face from that.
As for what exactly is stored and shared, that depends on the client and customer. As I mentioned before, we always work according to our clients’ requirements. Some are very strict with privacy and want zero retention of data. In our partnership with the Jumeirah Group, for example, we deployed the entire solution in our Asia cloud so all data stays within the Emerati state. We work with similar requirements to comply with European GDPR laws.
The bottom line is, it’s always up to the customer – they can decide how the data is used.
What makes Incode different from other biometric and identity verification companies?
Most of our technology, over 90%, is built in-house ourselves, not outsourced. Having tech built in-house gives us the flexibility and speed to move fast.
Much of the biometric market is still relatively new. Privacy and data laws are still changing. Owning the technology ourselves means we can be agile and build our solution based on what the customer wants and what current regulations specify.
Can you give us any hints as to the future of Incode? What other innovative projects are you working on?
I’m exploring ways we can use quantum computing to optimize scoring in some of our onboarding processes. In classical computing, we’re limited by two possible states of one bit in memory at one time (either 0 or 1). However, with quantum computing, which is governed by the laws of quantum physics, a single bit can be a superposition of both 0 and 1 at the same time. That way we can gain parallelization power for computation and potentially increase speed exponentially for the properly designed quantum algorithm.
With privacy policies evolving, we’re also exploring how identity verification and validation can happen completely on a user’s device. Much of our technology already operates on the edge so we can maximize speed and privacy, but we’re also looking at how we can build a completely offline solution.
Life at Incode is part of our blog series aimed at giving you a behind-the-scenes look at the people and culture of Incode. To learn more about our technology, visit our website. To learn more about being an Incoder, view our current available positions here.