Open-source actuarial engines are no longer a curiosity. modelx and lifelib run serious projection models, and uslib ships twelve U.S. life and annuity reference models. How ready is open source to step up to production work? Few, if any, carriers run it that way today. The math isn't the issue, and the engine isn't the limiting factor. This session asks what a carrier would need to build, validate, and govern to run an open-source model in production, rather than as a side check.

We'll go from proof of concept to production:

• Runtime and scale: how memory arithmetic works, and why portfolio rate-driven products such as UL and FIA change how a model can be split up and distributed.
• The completeness gap: what is the difference between a reference model and a production model? Some factors include rider inclusion, non-guaranteed element modeling, reinsurance, tax reserves, assumption management, and scenario generator integration.
• Validation and change control: how do we reproduce a prior quarter's results three years later during an exam?
• Governance: ASOP 56, ASOP 23, VM-31, and VM-G all need to be addressed. Your internal procurement, IT, audit, legal, and management staff will ask: Who patches the software? Who is on call on close night? What happens when an upstream release breaks something?

At the end, we will look at historical precedents from Linux, R, and QuantLib. After that, we'll explore steps a company might take. Those include: reporting and QA, choosing a specific block, running it in parallel, and finally making the open-source model the primary one. We'll also look at cost factors along the way. This is an honest engineering assessment, not advocacy.