Skip to main content
Back to challenges
Challenge

LlamaIndex & DeepSeek R1 Actuarial Risk Analysis Agent

Build a LlamaIndex RAG pipeline powered by DeepSeek R1 reasoning models to automate actuarial risk modeling and climate stress testing for regional insurers. Achieve 92%+ reasoning match against expert actuary benchmark evaluation standard.

Workflow AutomationHosted by Vera
Sign in free to start
Challenge brief

What you are building

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

Combine LlamaIndex advanced index querying with DeepSeek R1 reasoning capabilities for complex actuarial climate stress test analytics.

Delivery guide

How work is evaluated

Evaluation

Evaluates RAG retrieval context relevance and DeepSeek R1 reasoning fidelity on actuarial risk assessments.

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

Numeric Precision Check

Checks if estimated loss ratio is within expected range

Binary check

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

Dimension 2

reasoning_match_score

Step-by-step similarity to expert actuarial reasoning • target: 0.92 • range: 0-1

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

  • Construct advanced hierarchical indexes with LlamaIndex for complex tabular and text data

  • Integrate DeepSeek R1 reasoning models for multi-step math and actuarial reasoning

  • Implement climate stress testing models based on Asian Actuarial Conference guidelines

  • Optimize retrieval relevance through hybrid search and re-ranking

Resources and assets

Reference links and supporting material

Dataset notes

Synthetic dataset of ASEAN climate hazard projections and actuarial loss experience data.

Challenge setup

Evaluates RAG retrieval context relevance and DeepSeek R1 reasoning fidelity on actuarial risk assessments.

Challenge input

JSON containing scenario parameters and historical loss ratio docs

LlamaIndex

Data framework for structured and unstructured RAG indexing

DeepSeek R1

Advanced reasoning model for complex mathematical actuarial steps

Evaluated output

JSON containing reasoning_chain, estimated_loss_ratio, and solvency_impact

Evaluation checks
  • Checks if estimated loss ratio is within expected range
  • Step-by-step similarity to expert actuarial reasoning • target: 0.92 • range: 0-1
Configured evaluation evidence
  • Benchmark: MathQA / ActuaryBench
  • Reasoning Match Score target: 0.92
  • 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

Action Space
  • Llama Index · Required
  • FLAML · Optional
Policy Serving
  • DeepSeek R1 · Optional

Evaluation contract

  • Numeric Precision Check · Weight 1
  • reasoning_match_score · Weight 1

Metrics

  • reasoning_match_score · Target 0.92

Benchmarks

  • MathQA / ActuaryBench
  • general

Reference cases

  • 1 public reference case

Workflow tasks

  1. Task 1

    Input: JSON containing scenario parameters and historical loss ratio docs

    Output: JSON containing reasoning_chain, estimated_loss_ratio, and solvency_impact

    Execution harness: Python

Evaluation tools

  • LlamaIndex · Agent Framework
  • DeepSeek R1 · Models

Recipe freeze marker

No recipe freeze time is recorded.

Frequently Asked Questions about LlamaIndex & DeepSeek R1 Actuarial Risk Analysis Agent