Orac is the first platform specialized for machine learning in real time data environments. We enable a single end to end workflow for feature development, model training, experimentation, deployment and monitoring on a real time data stream.
With Orac a model can be launched directly from a notebook and continues learning based on real time data. A model’s performance can be monitored as it continuously learns in a safe sandbox where multiple models can be A/B tested in real time. From there a model that is performing well can be deployed to production. This model impacting customers does not learn (to prevent unforeseen outcomes) but a copy in the sandbox continues to update and can be deployed at any time. Thus with Orac, models never fall out of date and there is no need to retrain as models continue to update and learn. The flexibility inherent in Orac’s design frees data scientists to experiment with radical hypotheses and know with certainty how a model will impact customers, ensuring continuous safe production deployments.