Challenge

Build a Global Reinsurance Risk Aggregator with LlamaIndex and DeepSeek R1

Global reinsurers like Munich Re and Lloyd's require complex exposure modeling across thousands of multi-jurisdictional treaty documents. Build a retrieval-augmented generation pipeline using LlamaIndex and DeepSeek R1 to analyze treaty risk clauses. Achieve high precision context retrieval and exact clause compliance extraction across reinsurance treaty filings.

Machine LearningHosted by Vera
Challenge brief

What you are building

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

Construct an advanced LlamaIndex RAG system utilizing DeepSeek R1 reasoning to parse, chunk, and index global reinsurance treaty contracts.

Delivery guide

How work is evaluated

Evaluation

Evaluates LlamaIndex retrieval precision and loss retention calculation using DeepSeek R1.

Datasets

Shared data for this challenge

Review public datasets and any private uploads tied to your build.

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

retrieval_hit_rate

Required source clauses are present in retrieved contexts.

Binary check

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

Dimension 2

extraction_accuracy

Exact match accuracy on numeric retention amounts and listed exclusions • 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

  • Implement hierarchical chunking strategies using LlamaIndex.

  • Leverage DeepSeek R1 reasoning for complex legal exposure reasoning.

  • Build metadata-filtered vector indexes over multi-jurisdictional treaties.

  • Evaluate RAG retrieval accuracy with target hit-rate metrics.

Resources and assets

Reference links and supporting material

Dataset notes

Corpus of 50 synthetic reinsurance treaty contracts based on Munich Re and Lloyd's standard market clauses.

How this agent runs

Evaluates LlamaIndex retrieval precision and loss retention calculation using DeepSeek R1.

Challenge input

Natural language query specifying hazard type and region

LlamaIndex

Provides structural parsing and vector indexing components.

DeepSeek R1

Delivers extended reasoning over dense legal language.

Evaluated output

JSON containing retention_usd, exclusion_list, and target_clauses

Checks for
  • Required source clauses are present in retrieved contexts.
  • Exact match accuracy on numeric retention amounts and listed exclusions • target: 0.92 • range: 0-1
Proof of success
  • Benchmark: LegalRAGBench
  • Extraction Accuracy target: 0.92
  • 1 public reference case
Runtime evidence
  • Python execution harness
  • Python sandbox (unavailable on Versalist)
View technical recipe

Configured tools

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

Evaluation contract

  • retrieval_hit_rate · Weight 1
  • extraction_accuracy · Weight 1

Recipe state

This is a preview. The configuration can change before the evaluation recipe is locked.

Frequently Asked Questions about Build a Global Reinsurance Risk Aggregator with LlamaIndex and DeepSeek R1