Thought Leaders in Artificial Intelligence: Francisco Webber, CEO of Cortical.io (Part 6)

Sramana Mitra: So, it’s a combination of angel financing and a lot of bootstrapping using services. Now, you are at a point where you also have production-grade product systems. 

Francisco Webber: The big pivot was in 2019 when we decided that our current frameworks were stable and mature enough to boil them down into specific products. We were also lucky to get an investment round with a company called Xilinx. It turns out that using SPGA hardware allows us to speed up our algorithms. We were lucky enough to convince Xilinx that it makes sense to develop designs on FPGAs which speed up our main filter, for example.

On the business server, I can run 1 million emails per day through it easily. The same application would be used at a data center of a telco provider and they need to run a billion emails. The cost per filter and having this specialized hardware becomes relevant. This is a trend that you can observe in the AI market. People are trying to find more efficient hardware because things are starting to get expensive. To pay $10 million for a GPT-3 model is not possible for everybody.

By accident, we ended up betting on a good horse. We created a product prototype for our focus system as a complete appliance where you would get a OneU system. We partnered with Supermicro for that. We can deploy that functionality in an accelerated fashion. As we are not yet with the very big accounts there, it hasn’t become a burning thing, but it’s definitely something in the midterm. Just think of hate speech detection for Facebook. Just imagine how much computing power they would need to find out the subtle hate speech that could occur in any language. 

Sramana Mitra: So, Facebook doesn’t have any of these kinds of technology?

Francisco Webber: No. That was another lesson I learned very early. At the beginning of our technology, I went and knocked on the doors of Google, Apple, Facebook, and Twitter. What I saw was at that point they were not waiting for a little company from Austria to offer them something substantial. It is super hard to even trigger some awareness. I remember I met the product manager of Twitter at the time. They were desperately looking for a business model.

I said, “Look, it’s just an idea that I have, but what if you could provide business customers personalized feeds of your firehose only with tweets that are relevant to them. For example, you would provide Nokia with every tweet about mobile phones so that they could do analytics. If you try to do this with state of the art, it’s pretty expensive even for a company like Nokia to do this with 20,000 messages a second. You need to have a strong system to do this.”

I could and probably still can run a firehose filter on my notebook using our approach. As we were not well known and well established, it was impossible to get some attention. I believe that the more commercial success we show, the more attention we get. 

Sramana Mitra: What is the total amount of financing that you have put in so far?

Francisco Webber: Currently, it’s around 14 million euros. It is pretty lean. 

Sramana Mitra: All the 33 people in your team are all in Vienna?

Francisco Webber: No, we have our development team in Vienna. We have a customer success team in New York because our earliest large customers were in that area. Part of our management is on the West Coast. We are scattered between Europe and the US. 

This segment is part 6 in the series : Thought Leaders in Artificial Intelligence: Francisco Webber, CEO of Cortical.io
1 2 3 4 5 6 7