We use automated machine learning (AutoML) to save you money. - AutoML eliminates the inefficiencies, gruntwork, and mistakes involved in an AI project by trying all potential solutions in one program. It is like using a word processor vs. a typewriter, a smartphone vs. a flip phone, a text vs. a fax. Other ML programmers waste huge amounts of time (which you have to pay for) manually performing tasks we can do in seconds with the click of a button. This allows us to focus on the more important parts of your project while the automation handles the rest (data cleaning, feature extraction, algorithm testing, model deployment, etc.).
We use virtual teamwork in a special way. - After participating in machine learning competitions such as Kaggle, DrivenData, and AIcrowd, we noticed that almost all of these contests follow the same pattern. If, for, example the goal is to have AI look at a pizza and determine what kind it is (pepperoni, pineapple, cheese, etc.), entrants start off using a baseline model that might be 30% accurate. After a week, some people figure out improvements to get 50% accuracy. Then eventually, by the end of the competition, a few programmers reach 80% accuracy, almost always by building on the good ideas the other entrants used. It is somewhat of a collaborative effort.
We take this same concept and apply it to your project. Most ML consultants will get stuck at those low accuracy levels because they do all the work themselves, even if they are part of an agency or consulting company. Instead, we create an initial version of your machine learning model, and then have additional programmers try to improve it. Everybody works together as a team, sharing the same files, the same AutoML program, and the same cloud resources.
Over the past ten years, we have worked on typical machine learning projects as well as more complex areas such as algorithmic stock trading, neuroevolution, self-driving cars, genetic programming, reinforcement learning, and neural architecture search. But these are not things that can easily be put on the web, so we created a site at BoredHumans.com to showcase some of what we can do. It has over 40 fun AI pages that can be used online.
Below are some articles about how BoredHumans.com was built: