aktua risk
Request a demo
Private alpha · built for insurance teams

From raw insurance data to capital decisions.

Aktua connects data preparation, reserving, calibration, economic capital, reinsurance, and reporting inside one governed, local-first portfolio workspace.

Request a private demo Explore the workflow
Local-first Desktop execution
Versioned Published artifacts
Auditable Engine-grounded results
S Aktua Portfolio
North America P&C · 2026 plan
Economic capital

Portfolio result

Complete
Simulations 100,000 Production run
Correlation pairs 26,335 Validated
RI contracts 13 Applied in priority
Gross → net loss distribution Capital view
Gross Net
✓
Inputs pinned and verified 4 publications · immutable hashes · reproducible seed
just now
⌁
Correlation checkPSD valid · diagnostics passed
✦
AI briefing readyCalculated facts separated
ONE GOVERNED PORTFOLIO FOUR ACTUARIAL PRODUCTS ZERO BLACK-BOX CALCULATIONS
The Aktua suite

Four products.
One chain of evidence.

Work moves forward through explicit publications—not copy-and-paste handoffs. Every downstream model can identify the data, assumptions, and decisions it used.

01 ▦
Data studio

Datuak

Load, map, validate, explore, and publish canonical claims and policy data.

  • Schema-aware ingestion
  • AI-assisted mapping
  • Versioned data publications
02 ⌁
Reserving

Reserba

Build triangles, compare methods, select development, and approve indications.

  • Paid and incurred views
  • Method comparison
  • Controlled selections
03 ⌬
Calibration

Kalibra

Fit and approve frequency, severity, trend, and related model assumptions.

  • Fit diagnostics
  • Scenario comparison
  • Approved assumption sets
04 ▶
Economic capital

Simulat

Compose assumptions, dependence, catastrophe, assets, and protection into financial outcomes.

  • Monte Carlo simulation
  • Correlation and reinsurance
  • Flexible reporting
Designed as a system

Intentional handoffs, end to end.

The workflow preserves lineage from the first source table to the final capital report.

01
▦
Prepare

Canonical data

Profile, map, validate, and publish claims and exposure data.

data publication
02
⌁
Decide

Reserves and assumptions

Select development, fit models, and approve the basis for simulation.

approved basis
03
▶
Model

Portfolio outcomes

Apply dependence, catastrophe, assets, and reinsurance at production scale.

run manifest
04
↗
Explain

Reports and decisions

Build governed views, investigate drivers, and brief stakeholders.

decision record
Performance without theater

Large models belong on ordinary hardware.

Aktua is engineered for the workload that matters: realistic portfolios with hundreds of risk units, catastrophe events, dependence, and reinsurance—not a single-distribution toy benchmark.

Streaming artifacts and bounded working sets keep scale from turning into a memory emergency. The Rust engine uses the machine you have while preserving deterministic, reviewable outputs.

Review the benchmark methodology
Representative production benchmark Heavy CAT · correlation · reinsurance
Verified workflow
Iterations 100k
Risk units 230
Loss sources 883
Gross simulation parallel
Correlation bounded
Reinsurance priority-aware
Reporting queryable
↳

Benchmark inputs, fingerprints, and correctness gates are retained with the engineering record.

✦ Aktua analyst
Grounded mode
What changed after applying the active reinsurance program?
Calculated facts Engine verified
Simulations20,000
Active contracts13
Configured pairs26,335
Mean cession15.08%
Interpretation

The program reduces expected loss while preserving the portfolio’s principal tail drivers. The largest recoveries are concentrated in…

Planned actions require confirmation
AI with an evidence boundary

The engine calculates. AI explains.

Aktua’s analyst receives compact, calculated facts—not authority to manufacture results. Facts and interpretation remain visibly separate.

01
Grounded answers

Responses cite results produced by the actuarial engine.

02
Masked identifiers

Project names can be replaced locally before external AI use.

03
Human confirmation

Reruns and model-changing actions stop for explicit approval.

Reporting that stays connected

Move from result to reason.

Build repeatable views across dimensions, compare gross and net positions, inspect tail metrics, and trace every result back to its run.

CapitalLoss ratiosTail risk ReinsuranceAllocationsAudit
ExecutivePortfolioReinsurance
Portfolio capital reportRisk view by segment
Export ↗
42%Property
Property 42% Casualty 31% Specialty 27%
SegmentMean lossVaR 99.5
Property$854m$2.18b
Casualty$621m$1.56b
Specialty$477m$1.14b
Governance is product behavior

Designed to earn review.

Controls are built into the way work moves—not added as a report after the decision.

⌂

Local-first execution

Core model data and simulation run on the user’s machine by default.

#

Versioned publications

Data, assumptions, and decisions cross product boundaries with identity and hash.

↺

Reproducible runs

Settings, seeds, manifests, and fingerprints support repeatable investigation.

✓

Explicit approval

Material publications and AI-proposed execution remain human decisions.

Private alpha

Bring us the model that
should not run this fast.

We are working with a small number of insurance teams on serious actuarial workflows.

Start a conversation Founder-led demos · No source access required · Confidential by default