The In Conversation series is an opportunity for us to talk to the people that make the Victory Square Technologies family great. It’s an inside look at who they are, what they do and how they’ve become phenomenons in their respective fields.
Our latest sit-down is with Francis Doumet and Migel Tissera of Compression.Ai. Long-time technology industry vets, the two sat down with VST’s James Graham at Launch Academy in Vancouver to explain why compression is key and the surprising roles that their technology could play in the next few years.
James Graham: Gentlemen, let’s start with the most basic question of all, what is it that Compression.AI does?
Francis Doumet: We use machine learning to compress data down to as little as five percent of the original file size without compromising on the quality or the resolution of the original data.
We got started by processing images for websites and we produce jpegs that are less than half the size of traditional jpegs. As a result, web pages load up to forty percent faster and you get conversions and higher search rankings as a result.
Migel Tissera: Every second that you decrease the load time increases conversions by seven percent. If you’re an online retailer, a dating site or a real estate listing; the faster your page loads the better the business value. Compression helps with the storage component as well as the data transmission. When you compress, it’s less data to store and translate. By transmitting it faster, you get all these business values.
James: Is what you are doing specifically focused on the image compression market? Is it applicable to other file formats?
Francis: Image compression is the beachhead. Our first market is image compression, but that’s only the tip of the iceberg. Our algorithms can be applied to other formats, first and foremost video. It can also be applied to other formats like Lidar data for autonomous cars and medical imaging.
James: Is this the point of focus right now or are you attacking other things like video as well?
Francis: It’s the point of focus. Our product is essentially a tool for enterprise customers to increase the performance of their app or their online applications. For companies with websites, they can use our tool to compress all the assets on their website and have it load faster. For mobile games, they can compress all of the assets and get the download size smaller to decrease churn and download time.
James: What was the specific moment where you recognized that there was a pain point to be filled?
Francis: The thing with every compression technology out there today is that it’s just one formula that’s applied universally across the board to any file you try to compress. It’s very inefficient in a sense that it’s mediocrely good for every type of file you want to use. The good thing about machine learning is that you can look at the content on the file, adapt that formula so it’s no longer static and apply a different formula to different inputs. It makes it way more efficient and it helps us achieve our objective of making applications faster.
Migel: I think the “Aha!” moment was when we realized that we could use AI to supercharge existing algorithms and formats. We can create totally new ones which are really good too. That “Aha” moment was when we figured out we actually can improve these stock algorithms that have been widely adopted by fifty percent, just by rewriting the entire thing with AI.
James: What role does AI and Machine Learning play in Compression.AI??
Migel: Everything really. We are a true AI company. Everything we do is based on Machine Learning. The algorithms are AI-based. How we approach things is AI-based. When facing a new problem, all the solutions that we create are AI-based and based on neural networks.
James: How does Compression.AI increase my conversion rate? Is there an easy way to explain that?
Migel: The Internet works like it’s a postal network. You put an address in and all of the data packets come to you from a server. The larger the packet, the slower the delivery time. We compress all the assets so that the packets that get delivered are small so it comes to you really fast. When it comes to you fast, it’s obviously a great user experience. One of the key indicators in Google ranking is how fast the page loads. If you put the same content on two different URLs, the one that loads faster gets ranked higher. That’s how the SEO rankings come into play.
James: How did we get here? Where did Compression.AI come from?
Migel: The idea began when I was living in Australia where the internet speeds are really bad. In terms of the average speed ranking for the world, it’s not even in the top 40. I’m a big UFC fan. I would start watching a fight at 1080p streaming and the network quality would always degrade. The entire screen would downgrade into standard definition or 480P. I would start seeing pixels or worse, it starts buffering. Ten seconds later, the video is playing again and one of the fighters is on the floor. I don’t know what happened.
James: You just missed the knockout.
Migel: Exactly. I just missed the knockout of the night. To me, that’s not really a good user experience. I only really care about the two fighters in the middle and maybe the referee. If it’s a tennis match, it’s the same thing. I care about the tennis court and the two players. So I thought, what if we could adaptively compress? We keep the parts that matter to us in high definition and everything else gets pixelated once the network degrades. When I explained this idea to Francis he was like “Oh cool, I think we should look into that.” That’s how it started.
A video is a sequence of frames. We started with frames and we figured we needed to compress that first. That’s what we’ve done and we’ve done a great job at it, so the video is on the horizon.
Now, in regards to us. I moved here a year and a half ago. I have a background in engineering, mechatronics and most importantly, a Ph.D. in Machine Learning and AI. So I focus on those areas most in the company.
Francis: I have an engineering background from Stanford and worked in the Bay Area for a few years before going to business school at Wharton.
Migel: We manage the business side together for the most part, but he’s mostly in charge of that. We’ve got two software engineers working on projects that are very important for generating revenue.
James: Bringing on more software engineers suggests that you’re getting a good response from companies wanting to take advantage of your technology.
Francis: Yes, there are so many things we can do with compression that we want to try it all. We definitely have an interest in the technology where I think there’s still a ways to go to make it scale, make it proven. So we hired people to help us accelerate getting those products to market.
James: What differentiates you guys from the competition?
Migel: I think it’s the AI component that really differentiates us, these are true machine learning algorithms. We took the existing standards and rewrote them with AI. The other ones are just variants of the existing ones.
Francis: For a customer, they don’t necessarily care about the AI, they care about the performance,
Migel: We’re rendering images that are twice as small as a jpeg at the same quality. We’re 75% better than a prominent competitor. 50% better than the next, and 40% better than the one after that.
James: How does it feel to get props from Forbes?
Francis: Pretty good. It was a bit of a surprise, to be honest. We saw a spike in internet traffic and happened to see who was referring us.
Migel: The first Forbes article that was published was one we wrote and it got good traction. The second one we didn’t have anything to do with, they just picked us up.
(Editors Note: Francis and Migel have also been featured by NVidia since doing this interview)
James: What don’t we know about Compression.Ai that perhaps we should?
Migel: What people don’t see is how good the team is and how good the product and the vision is. It’s also the advisors and investors, I think it’s all of that combined.
Francis: Because our algorithms are so efficient it sometimes makes a lot of people skeptical. I think a lot of it is just because there’s a lack of knowledge about how the algorithms work. They don’t know about the core tech or the mission of the company. A lot of people think we’re just another image compression company and there’s not a crazy huge market. This is just the beginning, this just the beachhead. The mission for the company is to really accelerate online applications and that can go from websites all the way to autonomous vehicles, to satellites and into space.
Migel: We’ve connected with the special operations in the military here. They have a need for real-time communications. The word compression to some means that it’s just a savings thing. This isn’t really about just the savings, it’s a byproduct and it’s a good by-product to have. What it enables is what’s really important. What’s crucial is having someone in the field beaming back data to the control center for analysis or making decisions in military settings. That aspect of real-time communications, people don’t really realize that we will help bring that forward. It’s just going to take time and with the help of people like you, I think we can get that out.
James: Where do you guys see Compression.Ai five years from now?
Francis: If you look at the industry trend, bandwidth is always increasing and we have 5G coming. We want to be the company that enables new technologies that require more bandwidth. We’re increasing bandwidth because we need it. You can buy an 8K TV but there’s no 8K content, so that’s why they’re making the pipes wider. We want to be viewed as the company that is able to pump in more data through the existing pipes so new formats can be created.
You can learn more about Compression.Ai at https://compression.ai/
Francis Doumet and Migel Tissera are the minds behind Compression.Ai. We’re pretty sure Migel knows more about AI than Dana White.
James Graham is Victory Square Technology’s resident content creator. He doesn’t have a coffee problem anymore but he does have a half marathon problem.