Modeling Practices for Regulated AI
The Validation Framework That Satisfies Both Data Scientists and Regulators CFPB Circular 2022-03 made the regulatory position unambiguous: creditors using complex algorithms for …
•
25 min read
The Validation Framework That Satisfies Both Data Scientists and Regulators CFPB Circular 2022-03 made the regulatory position unambiguous: creditors using complex algorithms for …
Most data science projects do not fail because the algorithm is weak. They fail earlier. The business question is vague. The experiment is flawed. The team optimizes the wrong …
Most AI teams do not fail because the model is weak. They fail because the path from user feedback to production change is messy, rushed, and poorly governed. I have seen strong …
The 10 Best Practices That Separate AI Projects That Ship From AI Projects That Stall Managing AI development and deployment projects requires practices fundamentally different …
How to Audit AI Systems Beyond Approval and Into Real Operations Most organizations audit AI model approval thoroughly and audit AI model operations barely at all. They verify that …
From Manual Sampling to Monitoring 100% of Transactions GRC data scattered across disconnected systems. Compliance controls that depend on slow, human-driven processes never built …
Top Risk Scenarios and Controls That Actually Protect Your AI Project A risk register with 15 vaguely worded AI risks and a color-coded heat map is not a taxonomy. It is a …
Incident Types and Direct Loss Categories That Define Real Exposure for AI Projects Here is a question that reveals whether your AI risk program is mature or performative: Can you …
The Post-How to Keep AI Useful, Safe, and Accountable After Launch Most AI projects spend too much energy getting to deployment and not enough planning what happens next. That is a …
Most AI Impact Assessments Ignore Fundamental Rights Here’s the Category Taxonomy to Fix That Last year, I reviewed an AI impact assessment for a financial services firm deploying …