Video coming soon

What do you do?

Arthur is dedicated to bringing High-Performing AI Into Production Safely and Responsibly. Arthur is the platform we wished we’d had in previous roles, to provide much needed visibility into the large-scale systems we’d worked so hard to build. Our goal is to make every model observable, equitable, and auditable so that all AI/ML practitioners & stakeholders can understand and continually improve the operations of their systems.

How much does it cost?

Consumption based on volume of inferences


What’s a sample use case? Where can I learn from?

  • Recommendation & Decision-Making – Detect potential biases in your algorithms, quantify the importance of individual features to your predictions, and test hypothetical changes to your model by adjusting weights of certain features.
  • Regulatory Compliance – Quantify univariate and multivariate bias in decision- making models and explain model outputs — including financial loan decisions orhealthcare treatment plans — in nontechnical terms through exportable reports.
  • NLP & Computer Vision – Ensure consistency and monitor for data drift in text classification, image classification, or object detection information extraction pipelines, and identify the most important words or image areas in determining the outcome of your models.
  • Customer Churn – Improve your bottom line and understanding of your customers by pinpointing drivers of customer acquisition, churn, and lifetime value.
  • Process Automation – Track performance issues and changes in your data integrity (such as black swan events like COVID-19) that could affect the robustness of your process automations
  • Anomaly Detection – Set up intelligent alerts to notify you of impending data driftor disparate impact resulting from divergence from your training data.

Feature List

  • Data drift – Identify univariate and multivariate data drift in your ML models, both for tabular and unstructured data like NLP & Computer Vision models
  • Intelligent alerts – Set up intelligent alerts. Never miss an issue with your model inputs or performance
  • Performance monitoring – Track the performance of your models using research-backed metrics
  • Bias detection & mitigation – Identify biases in your data using univariate or multivariate segmentation, and take action to ensure fairness
  • Explainable AI – Infer how your models are making decisions using our Explainable AI features
  • Custom metrics – Choose which metrics work for you, including proprietary metrics from our renowned team of experts
  • Model diagnosis – Leverage monitoring and explainability features to diagnose and mitigate issues with your model(s)
  • Counterfactual explanations – Simulate “what-if” scenarios by adjusting features and observing changes in prediction impact
  • Exportable reports – Increase collaboration across teams with easy-to-digest exportable reports