AI & ML interests

AI & ML for Actuarial Science

Recent Activity

Organization Card

Intelligent Actuaries

An AI research lab for actuarial science.

Our goal is to advance the actuarial profession with AI. We publish our research and tools for everyone, and take on client work for a small number of selected clients.

GoogleLogoGIF-LARGE

This Hugging Face organisation holds the open data and demos: a synthetic experience-study dataset, and the studies and simulators you can run in the browser.

Website Research Client work


Two ways to work with us

Use our open work. Free for everyone: our research, the simulators that test it, and the desktop tools.

  • Working papers, each with its PDF and its Zenodo record
  • Four simulators that run in your browser, with nothing to install
  • Scelo 1.0, an AI workbench for actuarial work, for Linux, Windows and macOS
  • Docs, sample workspaces, and exports in Python, R and SQL

Become a client. We work with a small number of selected insurers, reinsurers and pension funds, by introduction only.

  • AI systems and research for reserving, mortality, capital, pricing and pensions
  • Run on your own systems, so your data never leaves them
  • The same specialists from start to finish, not a support queue
  • Every figure traced to its input, its model, and the rule it sits under

On Hugging Face

Acknowledgements

Our lapse and surrender study answers the Society of Actuaries (SOA) Research Institute and LIMRA 2023–2025 Term and Whole Life Lapse and Surrender Experience Study data request, which defines its scope and its VM-51 record layout. We gratefully acknowledge the SOA Research Institute and LIMRA for that request and for the experience studies it builds on. The data is synthetic, and the work is independent: it is not SOA or LIMRA data, and it is not affiliated with or endorsed by either organisation.


Tools

  • Scelo IDE: actuarial agents on your machine. They propose every change for your approval, engines compute every figure, and a ledger records each step. Download · In your browser · Docs
  • Community Lab IDE: a South African province of agents on published bases, with books, a burial society and paired experiments, on your machine. Download · Source
  • Scelo TUI (beta): the three-pane Scelo workstation in a terminal.
  • scelo for Python and R: the same models as a library (pip install scelo). Docs

Research

Working papers and preprints on how modern AI can serve the regulated work of insurers, reinsurers and pension funds. Each carries a DOI and live figures.

  • Ten Thousand Synthetic Policyholders: A VM-51 Lapse and Surrender Dataset, 2023–2025 (October 2026). Read · DOI
  • Community Lab IDE: A Province You Can Run (October 2026). Read · DOI
  • The Triset Table as a Count-of-Events Calculator (August 2026). Read · DOI
  • A Global Workspace for Actuarial Models (July 2026). Read · DOI
  • Scelo: A Private, Auditable Desktop Workbench for AI-Assisted Actuarial Analysis (June 2026). Read · DOI
  • A Coupled Power-Thermal-Cyber Framework for the Actuarial Pricing and Insurance of Hyperscale Data Centers (May 2026). Read · DOI
  • Nanoeconomics: Compounding Cooperation for Community Wealth (May 2026). Read · DOI

Simulators

The lab's work, live in your browser, with seeds and parameters exposed.

  • Scelo Atlas: price a hurricane against an insured book. A Holland (1980) wind field drives ground-up damage; the deductible, limit and share price it to net.
  • W(M, T, R): the Nanoeconomics community-wealth model, in single-community Monte Carlo and society-network modes.
  • SoftData: load a CSV or Parquet file and the workstation infers column types, runs a cleaning audit and stages the data for actuarial models.
  • Data Centers: the power-thermal-cyber data-center risk framework, with the paper's reliability, pricing, economics and OEP engines.

Contact


Where to find us


Intelligent Actuaries (Pty) Ltd is not licensed by the FSCA or supervised by the Prudential Authority. The regulators named in our work are the rules our clients answer to, listed for reference and not as an endorsement.

Built in South Africa 🇿🇦.

models 0

None public yet