Data used for model training can either be ingested via native NLP (natural language processing) based functionality that seamlessly feeds user comments into Curie or can be tagged on our platform and applied on historical datasets creating quality data which drives high-confidence predictions.
Curie offers both data scientists and non-technical business users the ability to select algorithms, train models, fine-tune parameters through a friendly UI to predict reasons for data quality issues and to detect anomalies instantly—saving valuable time spent researching root cause of data quality issues. Models are pre-configured with finely-tuned hyperparameters based on our IP and on our intimate knowledge of financial products.
The platform accommodates externally-built models and provides a robust framework for model management and governance with full auditability and explainability of model’s decisions—a feature of increasing importance to internal risk teams and regulators alike.
Historical performance data points and self-learning and improvement capabilities make the platform an easy choice for automation practitioners.