Skip to main content
Back to challenges
Challenge

OpenAI Agents SDK Reinsurance Portfolio Risk Engine

Reinsurance executives need autonomous agents that analyze multi-billion-dollar portfolio concentrations across global natural catastrophes. Build an OpenAI Agents SDK system in Replit Agent to evaluate cumulative catastrophe exposure and calculate Tail Value-at-Risk across reinsurance tranches.

Machine LearningHosted by Vera
Sign in free to start
Challenge brief

What you are building

The core problem, expected build, and operating context for this challenge.

Orchestrate OpenAI agents equipped with custom tool calling to calculate portfolio TVaR and recommend retrocession transfers.

Delivery guide

How work is evaluated

Evaluation

Evaluates accuracy of calculated 99.5% TVaR figures and appropriateness of recommended retrocession limits.

Datasets

Challenge datasets

Public datasets, plus the private datasets you have access to.

Evaluation rubric

How submissions are scored

These dimensions define what the evaluator checks and which criteria separate a passable run from a strong one.

Dimensions
2 scoring checks
Binary
2 pass or fail dimensions
Ordinal
0 scaled dimensions
Dimension 1

tvar_tolerance

Verifies TVaR numerical calculation is within 1% relative error

Binary check

This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.

Dimension 2

calculation_error

Relative percentage error against exact mathematical result • target: 0 • range: 0-0.05

Binary check

This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.

Learning goals

What you should walk away with

  • Utilize the OpenAI Agents SDK to build tools for parsing portfolio treaty documentation

  • Implement Monte Carlo Tail Value-at-Risk (TVaR) routines called directly by agent functions

  • Configure Replit Agent workflow for interactive reinsurance portfolio optimization

Resources and assets

Reference links and supporting material

Dataset notes

Reinsurance treaty portfolio table detailing regional hazard exposures across US Hurricane and EU Windstorm.

Challenge setup

Evaluates accuracy of calculated 99.5% TVaR figures and appropriateness of recommended retrocession limits.

Challenge input

JSON exposure vector and loss distributions

OpenAI Agents SDK

Provides official tool execution patterns and agent management

Replit Agent

Facilitates rapid agent iteration and web deployment

Evaluated output

JSON containing calculated_tvar, recommended_retrocession, and reasoning

Evaluation checks
  • Verifies TVaR numerical calculation is within 1% relative error
  • Relative percentage error against exact mathematical result • target: 0 • range: 0-0.05
Configured evaluation evidence
  • Benchmark: Actuarial-Risk-Eval
  • Calculation Error target: 0
  • 1 public reference case
Configured runtime
  • Python execution harness
  • Python sandbox (unavailable on Versalist)

Summary shows the first task, up to three tools, and up to three evidence items. Open the technical recipe to view all configured items.

View technical recipe

Configured tools

Environment
  • Replit Agent · Optional
Action Space
  • OpenAI · Required
  • Replit Agent · Optional
  • LocalAI · Optional

Evaluation contract

  • tvar_tolerance · Weight 1
  • calculation_error · Weight 1

Metrics

  • calculation_error · Target 0

Benchmarks

  • Actuarial-Risk-Eval
  • general

Reference cases

  • 1 public reference case

Workflow tasks

  1. Task 1

    Input: JSON exposure vector and loss distributions

    Output: JSON containing calculated_tvar, recommended_retrocession, and reasoning

    Execution harness: Python

Evaluation tools

  • OpenAI Agents SDK · Agent SDK
  • Replit Agent · Development & Execution Environment

Recipe freeze marker

No recipe freeze time is recorded.

Frequently Asked Questions about OpenAI Agents SDK Reinsurance Portfolio Risk Engine