David Yu, FSA · uModel
Products, assumptions, and regulations have always been modeled in different ways. uModel expresses them one way, so you can compose complex models from simple concepts.
Transparent, distributed, cloud native, and AI agent empowered.
16,000
seriatim contracts, per-contract solves
7 × 50
scenarios × projection years
2m 13s
full run on a desktop. Finding, 1 Sep 2026
The question
You can see the answer in your head: this benefit, that basis, this result. Then come the code, the run order, the timing, the aggregation and the reconciliation. That gap is where the years went: year after year of modeling, most of it spent on technical work rather than actuarial thinking.
The answer
Products, assumptions, and regulations come in endless variety, but underneath they rest on a small set of actuarial ideas. Working from those ideas rather than from each case's surface is what let us find one common way to express them all. Three things follow from it:
Do just that. Products are composed from building blocks, and you write only the economics of the case. Term, universal life, and deferred and payout annuities today.
Each contract is its own unit of work, so a run uses every core of the machine.
Inputs and outputs are well defined, and every number traces back to them. That gives an assistant something real to read.
No more struggling with new product features and regulation.
Built to distribute: every core engaged. Desktop-proven today, cloud native by design.
Model with an agent that understands the input you’re working on, advises how to do it, and helps check for errors.
A new way of working: you decide what the model should do, and an agent beside you advises on each input and helps catch errors as you go.
Development maturity · October 2026 · stage reached, in quarters
| Modeling challenge | Architecture | Core Build | Applied Validation | Pilot Readiness |
|---|---|---|---|---|
| Compose arbitrary contracts | ||||
| Comprehensive, transparent output | ||||
| Quantify income for cohorts and plan groups | ||||
| Multiple currencies in one model | ||||
| Nested projection | ||||
| Composable reinvestment strategy | ||||
| Planned: hedging strategy you define | ||||
| Cloud cost | ||||
| AI agent integration |
each dot = one quarter of the stage complete
Findings
Each one states what was run, what came out, and what it does not prove.
1 Sep 2026
16,000 contracts · 7 scenarios · 50 years · 2m13s
Read the Finding →
In progress
Next Finding. Published here when it's done.
Planned
Follow the build to get it first.
I'm David Yu, FSA. I've spent decades building actuarial models for life insurers, and taught myself software engineering to build the system I kept wishing for.
More about me →The engine runs today; the platform around it is being built in the open. Occasional notes, no spam.