Actuarial Modeling,
Composable.

The mission of uModel is to address the following modeling challenges.

Development maturity — September 2026

Development stage reached for each modeling challenge, in quarters
Modeling challenge Architecture Core Build Applied Validation Pilot Readiness
Compose arbitrary contracts Complex product features and assumptions conventionally mean new model code. Riders, crediting variants and benefit structures multiply faster than a codebase can absorb them. 4/4 4/4 4/4 3/4
Comprehensive, transparent output Results usually arrive pre-aggregated, so assembling a financial statement — or anything the designer did not anticipate — means going back to the model instead of to the output. 4/4 4/4 4/4 3/4
Quantify income for cohorts and plan groups Income has to be attributed — including investment income — across cohorts of contracts or groups of plans, because that grouping is what several regulatory frameworks are built on. 4/4 4/4 2/4 1/4
Multiple currencies in one model Business written in different currencies has to be modelled in its own currency and still aggregate, which means conversion has to be automatic rather than a reconciliation step. 4/4 4/4 2/4 1/4
Nested projection A projection that must value something at every future point needs another projection inside it — and the shape of that inner run has to be a design choice, not a fixed behaviour. 4/4 4/4 2/4 1/4
Composable reinvestment strategy Buying and selling across asset types is usually hard-coded. Expressing a strategy as composable steps is what makes it changeable without rebuilding the model. 4/4 4/4 2/4 1/4
Hedging strategy you define Hedging programmes differ in what they target and what they trade — delta only, a full variable-annuity programme, or an ALM programme by key-rate tenor. Conventional tooling fixes those choices in code, so the programme bends to the system. 4/4 2/4 0 0
Cloud cost The usual way to cut a compute bill is to approximate — compress, cluster, replicate — which spends accuracy to buy time. Spreading work across threads and machines does not. 4/4 3/4 2/4 2/4
AI that learns a model Understanding an unfamiliar model means reading it. An assistant that has actually learned a specific model can answer questions about it, rather than just follow instructions. 3/4 2/4 1/4 0
0 1/4 2/4 3/4 4/4
Architecture
The general solution is designed, including how it connects with the rest of the platform
Core Build
Implemented, and supported by internal unit, integration, numerical and regression testing
Applied Validation
Exercised on realistic actuarial cases and reconciled against independent expectations or existing results
Pilot Readiness
Documented, usable, performant, deployable and stable enough for work with an external partner

Follow the build.

uModel is in active development. The engine runs today; the platform around it is being built in the open.

Occasional notes on what's being built. No spam, and we won't share your address. Privacy.