Peril risk, made portable. Underwriting, made defensible.

Atlas United is a deterministic peril modeling company. We grade 282.9 million U.S. addresses across 54 perils through a single audit-ready API — calibrated against two decades of validated severe-event observations and a sixteen-day forward layer. Sub-100 ms cold latency, full methodology trace per response. Patent pending under USPTO Application #64/070,786.

Why this exists

For thirty years, the way insurance prices peril risk has come from a small handful of vendors selling proprietary catastrophe models that nobody outside the vendor can fully see inside. Carriers paid seven figures for the privilege. Regulators got vendor whitepapers and trade-secret disclaimers. Underwriters got a number and were expected to trust it.

Atlas United was built on a different premise: that a deterministic peril model can be transparent, address-level, and defensible — and still be the most accurate way to price risk. One model. Fifty-three perils. 282.9 million addresses. Zero null cells across the matrix. A single API call returns the grade, the score, the reason code (patent claim 30), the model version, the calibration vintage, and a cryptographic receipt of the response.

We are pre-revenue on the enterprise side, with USPTO-Filed core technology (the Address-level Event Probability Model — AEPM, USPTO Application #64/070,786) and a live production stack. Sub-100 ms cold latency on the fast endpoint. Live demos for carrier portfolio scoring, real-estate disclosure widgets, and cryptographic verification receipts.

Operating principles

Six things we don't compromise on.

Methodology over marketing.

Every score must be traceable to a documented model component with version, calibration vintage, and confidence interval. If we can't show the receipt, we don't ship the number.

Per-call, never per-seat.

Pricing should track value delivered, not the ability of procurement to extract rent. We will never charge for adding a teammate to a project.

Transparent peril modeling.

No black-box blending, no proprietary aggregator math. Each peril layer is independently calibrated and documented so any underwriter can audit the chain.

Document the failure modes.

Every model has uncertainty. We publish ours — confidence intervals, geographic blind spots, known biases — so customers can scope our data to their risk appetite.

Cryptographic receipts, by default.

Every response carries an Ed25519 signature, content hash, and Merkle root over the model layers used. Audit-ready out of the box.

Boring infrastructure, ambitious science.

The API should be the least surprising part of the day. The model behind it should be worth filing a patent on.

Where we are today

A working product, honestly described.

282.9M

U.S. addresses indexed across 54 perils

22 yr

Validated severe-event catalog calibrating the model

16 day

Forward probability layer with sub-hour refresh

<100ms

Cold p95 latency on the fast endpoint

Stage

Pre-revenue. USPTO-Filed. Raising pre-seed.

Atlas United is solo-founded and bootstrapped. The full peril modeling pipeline, the partitioned address cache, the verification-receipt scheme, and the live admin demos were built and shipped over the past several months. We are now opening enterprise pilots and a small pre-seed round to fund go-to-market and the first engineering hires. We will list institutional backers on this page when we have them — not before.