Currently in development — Early Access 2026

Actuarial reserving,
driven by agents

ReserveLab uses reinforcement learning to automate and augment P&C loss reserve calculations — from chain-ladder parametrisation to stochastic cashflow profiles — for global insurers and reinsurers.

Join the Waitlist View on GitHub
Loss Development Triangle · 10×10 · Auto BI
6+
Lines of business
4
Reserving methods
99.5%
VaR for Solvency II
2
Products
// The problem

Reserving is still
done by heuristics

  • Actuaries select development factors without formal statistical criteria — relying on experience and judgment alone
  • Model diagnostics (residual tests, calendar year effects) are rarely applied systematically
  • Tail extrapolation curves are chosen manually, with little sensitivity analysis
  • Cashflow profiles from reserves are rarely validated — yet critical for ALM and Solvency II capital
  • No formal process for documenting parametrisation decisions to regulators
Today
  • Development factors selected by judgment — no statistical basis
  • Model fit rarely tested — residual patterns go undetected
  • Tail assumptions chosen manually, sensitivity rarely quantified
  • Reserve uncertainty estimated informally or not at all
  • Cashflow profiles not validated against model assumptions
  • Parametrisation decisions undocumented — hard to defend to regulators
With ReserveLab
  • RL agent selects optimal parameters — formal, reproducible, auditable
  • Full residual diagnostics run automatically at every step
  • Tail curves fitted and compared by AIC/BIC — best fit selected
  • Reserve distribution with VaR 99.5% and TVaR for Solvency II SCR
  • Quarterly cashflow profiles validated against stochastic simulations
  • Complete audit trail — ready for board, external auditor, and regulator
// Products

Two ways to work
with ReserveLab

Whether you want full automation or prefer to stay in control, ReserveLab has a product for your workflow.

Autopilot
ReserveLab Autopilot
For scale and efficiency

A reinforcement learning agent that autonomously selects reserving methods, parametrises development factors, fits tail curves, and produces a fully documented reserve estimate — ready for actuarial sign-off.

  • RL agent optimises parametrisation across all accident years simultaneously
  • Formal residual diagnostics embedded in the reward signal
  • Stochastic cashflow profiles with quarterly granularity
  • VaR 99.5% and TVaR for Solvency II SCR
  • Full audit trail — suitable for regulatory review
Studio
ReserveLab Studio
For actuaries who want control

An interactive dashboard where actuaries set parameters themselves — with the RL agent as a second opinion, formal diagnostic checks at every step, and professional reporting output.

  • Interactive LDF selection with live diagnostic feedback
  • Tail curve fitting (Hoerl, Sherman, spline) with AIC/BIC comparison
  • Paid vs incurred triangle comparison — automatic discrepancy flags
  • Bootstrap and stochastic simulation for reserve distributions
  • Quarto PDF report — for management and regulators
// Platform

Built on open-source foundations

The ReserveLab platform is built from modular open-source components — available on PyPI for the actuarial community.

01 / reservelab-sim
Triangle Simulator
Synthetic P&C loss triangle generator. Supports paid and incurred, multiple lines of business, hierarchical scenario parametrisation, and configurable granularity.
02 / reservelab-methods
Reserving Methods
Chain-Ladder, Bornhuetter-Ferguson, Cape Cod, and ODP-GLM. Tail extrapolation via Hoerl, Sherman, and spline curves. LDF smoothing with credibility weighting.
03 / reservelab-diagnostics
Model Diagnostics
Formal residual tests: Pearson, Runs, Spearman trend, Mack calendar year effect, Levene homoscedasticity. Structured pass/fail output with p-values.
04 / reservelab-agent
RL Agent
PPO-based reinforcement learning agent trained on synthetic triangles. Reward function combines reserve accuracy, residual quality, cashflow precision, and YoY stability.
05 / reservelab-api
REST API
FastAPI endpoint accepting paid and incurred triangles, returning reserves, cashflow profiles, VaR estimates, and full diagnostic JSON — integrates with existing actuarial systems.
06 / reservelab-studio
Interactive Studio
Streamlit-based actuarial workbench. Upload triangles, adjust parameters interactively, inspect diagnostics visually, and export professional PDF reports via Quarto.
// Early Access

Be first to know

ReserveLab is currently in development. Join the waitlist to get early access, updates, and be part of shaping the product.

✓  You're on the list. We'll be in touch.

No spam. No commitments. Unsubscribe anytime.