ModelOp gives Data Scientists flexibility while allowing them to demonstrate the value of their work
Data scientists need a repeatable method to prepare models for production using the tools in their personal toolkit and from the workbench of their choice. Including ways to monitor drift of production data, to easily backtest models against responses from previously scored data, and easily retrain the model as necessary.
Flexibility of Language, Work Bench, IDEs, and Development Tools
Regardless of how the model was developed, every model needs to be tested, deployed into production, backtested, and retrained. We can help integrate this cycle with any language, work bench, IDE, and development tool a data scientist wants to use.
Monitor statistical assumptions
A data scientist only has what training data is available at the time the model is created. If new production data has statistical properties which violate the assumptions made at the time of creation, the data scientist should be alerted so that the model can be retrained or assumptions updated.
Transparency of model efficacy
Data scientists should have the opportunity to demonstrate how effective their models are in impacting their business’ bottom lines. One good way to demonstrate this is to routinely backtest the model as new responses to previously scored data becomes available.
Compare and contrast models
The joy for every data scientist is deepening their understanding of their data and developing new, better models which reflect that better understanding. There should be a way to easily demonstrate that a new model is an improvement over its predecessor.
How Your Work Will Be Transformed
Data scientists are able to develop models in any of the market leading workbenches, for example a Jupyter notebook, quickly iterating through research on the historical data set, and then quickly operationalize the model. ModelOp sets-up a repeatable system for backtesting as well as a dashboard business users could use to interact with the model.
How ModelOp Is Unique
Models are unique as software assets in that they are developed by teams with unique mathematical, statistical, machine learning, and AI knowledge. ModelOp allow teams across a business to modularize these assets into reusable pieces as well as demonstrate their efficacy to business owners.