Strategy

May 22, 2026

9 min read

Buy, Build, or Pre-Fab: A Third Path Out of the Enterprise AI Conundrum

Buying drags you through a six-month sales cycle. Building leans on one or two heroes and stalls at the cold start. A third paradigm — pre-fabricated projects — bypasses both, and the incremental time-to-market usually pays for the pre-fab on its own.

Reg Environment

May 15, 2026

12 min read

The 2026 Regulatory Environment for Models and AI: A Practitioner's Map

SR 26-2 has landed, the EU AI Act's high-risk obligations are biting, and NIST AI RMF and ISO/IEC 42001 are quietly becoming the lingua franca of AI governance. A field guide to what each one demands, who it applies to, and where Forli plugs in.

Observability

May 8, 2026

8 min read

Watch Yourself: Monitoring and Observability Should Be Autonomous, Automatic, and On by Default

Most ML systems get observability bolted on after the first incident. That's backwards. Monitoring should be a property of the model from the moment it's born — autonomous, automatic, and never optional.

Engineering

May 1, 2026

10 min read

Uncle Bob Was Right: Clean Architecture Belongs in Data Science

Clean Architecture isn't a relic of enterprise Java — it's the missing discipline that turns brittle ML pipelines into systems that survive contact with reality. Here's why we apply it to data science work, with concrete examples.

MLOps

April 24, 2026

9 min read

The Quiet Crisis: Why Enterprise ML Stalls Between Notebook and Production

Most enterprises celebrate a model that works in a notebook and ignore the eighteen-month death march to deploy it. Here's why that gap is the single most expensive blind spot in modern data organizations — and what to do about it.