Causaris builds causal models and Bayesian inference systems for institutions whose decisions face scrutiny from regulators, courts, and the public. Every estimate ships with its uncertainty. Every claim ships with its assumptions. Public-interest analyses are published in full, so anyone can check them; client work stays private, held to the same standard.
Causaris builds causal models and Bayesian inference systems for institutions whose decisions face scrutiny from regulators, courts, and the public. Every estimate ships with its uncertainty and every claim with its assumptions. Public-interest analyses are published in full — data, code, decision thresholds — so anyone can check them; client work stays confidential and is built to the same standard.
A conclusion that cannot be re-derived is a conclusion the reader is asked to take on faith. Publishing the data, code, and thresholds invites attack, and public contest is the strongest quality control there is. Several published analyses have been contested; to date, no error in the data, code, or inference of any of them has been demonstrated.
Machine learning earns its place in pipelines — feature extraction, anomaly triage, scale — but it never gets the last word on a conclusion. No finding rests on a model whose reasoning cannot be articulated, inspected, and challenged.
Causaris is led by Niko Gamulin, PhD in machine learning, with 15+ years building statistical and machine-learning systems in production, peer-reviewed publications, and a CRR3-oriented realized-price benchmark in production use at a banking group.