Company Performance Metrics
Piotr Plonski: Founder
MLJAR is a platform for rapid prototyping, developing and deploying machine learning models. Yeah! Rationale behind MLJAR:
(i) the universal learning algorithm does not exist, that's why to find good machine learning model you have to check many different algorithms - MLJAR makes it easy, don't worry!
(ii) learning algorithms have
set of hyperparameters that needs to be tuned for a problem - with MLJAR each selected algorithm is tuned to your dataset
(iii) training many different algorithms can take a lot of time - that's why MLJAR is doing this on many machines in parallel to give you results as fast as possible
(iv) there is a pain when you train a model and realize that you dont saved them, so you need to start training again - it will never happen with MLJAR, all models are always saved. What is more, we save models during training so you can check partial results and stop straining when you want. (We also store all models predicitions produced during training, so ensembling is easy, yeah!)
(v) sharing results with cooperators is a pain, you fill some spreadsheets and dump ROC curves into figures and attach them in email message - no more, with MLJAR you can easily share your results with others (even with whole world, by public projects)
(vi) to deploy your model you don't have to be a cloud hero, you just select algorithm, click and MLJAR provides you a REST API for your model
