Under the hood

Fast enough to quote live.
Built to own and defend.

The detail a technical evaluator wants to see: how a full-scope best-execution quote returns in under two seconds, how every result can be replayed field by field, and why behavior is configured as data, not code — the architecture of an engine a team can take ownership of.

<2s
per loan, full-scope
interactive best-execution
<49s
for a 250-loan batch in
Bulk Pricer — same engine, at scale
490+
stored procedures behind
the engines (by policy)
96
foreign keys enforcing
referential integrity

Throughput from the platform's own processing — measured, not projected.

The design

Six decisions a technical evaluator will want to see

Provider + executor-chain engine

The Program Engine is a Chain of Responsibility assembled by a factory — each executor handles one stage and passes an accumulating scenario forward. Program inheritance (shell programs drawing sections from root programs) is resolved entirely in the data layer, so the executors stay inheritance-unaware.

Stored-procedure data layer, tuned for volume

All data access is stored-procedure-based via Dapper. The ratesheet base grid is the platform's largest dataset — investor rate sheets push well over a million rows per day — so the pricing layer is deliberately tuned to evaluate only the records that can apply to a given loan.

Temporal loan history + audit replay

Loan Pricer owns the loan tables: a parent header plus nine system-versioned temporal child tables that retain full per-facet history automatically. Every run serializes its scenario to an executor log, which the audit screen re-opens to replay eligibility in a no-write mode for repeatable, field-by-field diagnosis.

Credit analysis feeds eligibility

For Non-QM scenarios, the borrower's credit events and mortgage lates are graded, and the resulting credit profile filters eligibility to the credit-approved programs before pricing ever runs — credit shaping the result, not bolted on after it.

The whole instrument, priced

Beyond rate and price, the engine resolves each option's ARM margin and rate caps and its prepayment terms — so an adjustable-rate loan is priced as the instrument it actually is. The kind of domain depth only an industry veteran would build in.

Behavior is configured, not coded

Guidelines and overlays are authored visually as data and bound to live loan fields at evaluation time — one authoring tool drives both engines. A lender reshapes eligibility and pricing behavior through configuration, with no deployment.

🔎

Diligence a closed SaaS can't offer

Loan Pricer is backed by the same independent, file-level Forensic Code Review as the rest of PPE1 — the kind of technical due diligence a team gets to read before it owns the engine. Consistent data access by policy (stored procedures via Dapper), enforced referential integrity, and a fully auditable eligibility decision mean what you evaluate is what you run.

The full architecture is in the data room.

Qualified parties get the Forensic Code Review and a guided technical walkthrough of the live engine.