The Machine Learning Assistant

Kortical claimed top spot at the Schroders Datathon vs other AutoML and data science teams. Data scientists using the platform have also got top scores on Kaggle but the results we’re most proud of are our customers having wins like 50x productivity boost, +56% revenues, reducing wastage 54%.

Kortical’s flexibility and control enables it to get the same high quality results when dealing with an imperfect reality.

Kortical is designed by data scientists, for data scientists. It doesn’t try to dumb down data science, but gives access to the code and insights putting the data scientist firmly in the driver's seat. Kortical is transparent and provides control over every aspect of the machine.

Kortical does automatic feature engineering and data cleaning, but more importantly it gives the data scientist the tools, libraries and insights to create powerful features of their own.

While you can easily create combination features with Kortical and there are templates in the platform to do so, the automatic feature generation does not do this yet, so we have marked it down slightly for that.

Kortical not only gives stats about your data, but uses machine learning powered insights to help you identify key factors in your dataset and speed up the understanding of your data.

One example is automatic feature selection, to detect which features can be safely dropped from your analysis, which is especially useful for large datasets. We created a best in class algorithm and provide not only the functionality but the paper on how it works so you can verify its output.

Most platforms use an imposter model like a random forest to get the feature importances. This does not really explain how your model works but gives a general idea of what’s important in the dataset. Kortical explains every model from linear models to Deep Neural Networks with NLP, with no shortcuts. This is much safer for really understand how your model works, reducing surprise behaviour and satisfying regulators.

Datarobot has more ML Ops features, but Kortical’s reliability is second to none, hence Kortical being chosen to power the NHS Blood Supply Chain and it is very easy to use and manage. Amazon and Google provide all the raw components needed, but there is much more do it yourself in their workflows.

Kortical also supports zero downtime hot swapping of live models, multiple environments for staging, testing and production as well as options for full hardware isolation.

Datarobot has some apps that if you want exactly that app you can use out of the box, but that is not particularly flexible, doesn’t provide much control and is not transferable if you want to move off Datarobot.

Kortical has taken a different approach providing app templates and the source code to work from as well as one line commands to get them live. These all use standard technology and are transferable outside the Kortical ecosystem. Having access to the code means that you can make whatever you want and Fortune 500s have full machine learning microservice architectures built using Kortical apps.

Kortical is designed to do what it does in the best way possible, but platforms like AWS and Google have myriad other parts to their solution that are not available in Kortical. The best way to do speech intent detection is AWS or Google speech to text + Kortical. So it’s not a case that you just have to use one.

With Kortical’s focus though comes a level of ease of use and speed.

We are super passionate about what we do and love to connect with our users. We have a slack community you can join here. You can ask a question directly to the CEO, CTO or any data scientist, vote for the features you would like to see, leave feedback in app or get involved in feedback sessions to help shape the future of the platform.

We really want to create the best data science assistant that you love using because it enhances your abilities and doesn’t restrict your ability to create the solutions you want or hide away important details.