Octeract was founded in 2017 by Nikos Kazazakis and Gabriel Lau.
Many years ago, we took a long hard look at the state of nonlinear optimisation software. It was pretty easy to see that humanity’s ability to solve linear optimisation problems has improved by millions of times over the last 25 years, whereas our ability to solve nonlinear optimisation problems has improved by a few dozen times at best, despite unimaginable amounts of research money.
As such things go, you can either be a hypocrite and complain how other people suck at doing something, or you can try to do it better yourself.
Our first attempt at solving this problem was Octeract Engine, which improved things by a few hundred times, especially with HPC. Even though it’s arguably the best nonlinear solver in the world, by our standards even that level of performance is pretty pathetic. We need millions of times.
Luckily, building Octeract Engine was great training. We came to understand that, even with stellar implementation, the fundamental limiter was the underlying technology stack – there was something missing. In response, we created a complementary technology, Octeract Neural. Neural implements a new theoretical framework we created for nonlinear optimisation called the Deep Evolution Metric, which allows us to solve general optimisation problems using machine learning. Neural technology is cutting edge so be patient as we improve it. It comes with its own challenges and limitations, but it doesn’t suffer from the curse of dimensionality like conventional technology does, so we can’t wait to see how its solving power evolves over time.
In a nutshell, our singular focus is to push the technological frontier. We value knowledge, information, talent, hard work, and results, with emphasis on the latter two.
Assuming that you have nonlinear problems that need solving, cutting edge technology will give you an edge. The clue is in the name.
When it comes to collaborations, we value building close, lasting relationships with the people we work with. This is typically achieved by as much face time as is necessary to get the job done, and, of course, getting the job done.
It doesn’t matter what it is, we’ll make it work. This is sometimes mistaken for confidence, but it’s not. It’s observation.
If you work on nonlinear optimisation, this is the best research tooling you can get. Using old technology for research can lead you down the wrong path, and you might spend years solving a solved problem, so be careful and make up your own mind.
We have special deals for research groups, as well as all sorts of experimental options and customisation for your research needs.
Contact sales to find out more!
We always welcome talented, hard-working people to apply. This is the most interesting job you’ll ever have, and you’ll get to work with the best technology and the best people.
Be warned though: work-life balance is not a phrase in our vocabulary. We work very hard and because of that we know our shit. If you want to work with us, so will you.
If you tried the engine and you were like “bruv wtf, this solver is slower than CPLEX/GUROBI, you said it was the best”, and when I ask you what you tried to solve the answer was an (MI)LP, this section is here specifically for you.
You are almost certainly entitled to a refund to your MSc/PhD, so maybe you should get cracking on that.