Skip to main content
Back to challenges
Challenge

COSMIC Intelligence Sharing Graph Agent with LlamaIndex and Pydantic AI

Singapore's COSMIC platform requires real-time detection of Money Laundering and Terrorism Financing across financial institutions. Construct a multi-node Knowledge Graph RAG application using LlamaIndex and Pydantic AI to validate cross-bank transaction disclosures. Achieve strict graph query accuracy and verifiable regulatory alert flagging.

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.

Build a collaborative AML/TF knowledge graph extraction and reasoning agent using LlamaIndex and Pydantic AI.

Delivery guide

How work is evaluated

Evaluation

Evaluates entity resolution accuracy and risk assessment speed across financial nodes.

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

exact_entity_match

Ensures all target suspicious entities are correctly identified.

Binary check

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

Dimension 2

f1_score

F1 score of entity threat identification. • target: 0.95 • 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 a LlamaIndex PropertyGraphIndex for entity-relationship extraction from banking disclosures

  • Define type-safe Pydantic AI models for AML risk classification

  • Implement multi-hop reasoning algorithms across connected financial entities

  • Evaluate precision and recall of suspicious activity report (SAR) triggers

Resources and assets

Reference links and supporting material

Dataset notes

Synthetic multi-bank dataset mimicking Singapore COSMIC transaction sharing logs.

Challenge setup

Evaluates entity resolution accuracy and risk assessment speed across financial nodes.

Challenge input

JSON graph adjacency list of transactions

LlamaIndex

Required assigned framework for graph RAG.

Pydantic AI

Assigned agent tool for type-safe execution.

Evaluated output

JSON list of flagged high-risk entities

Evaluation checks
  • Ensures all target suspicious entities are correctly identified.
  • F1 score of entity threat identification. • target: 0.95 • range: 0-1
Configured evaluation evidence
  • Benchmark: FinAML-Graph-Bench
  • F1 Score target: 0.95
  • 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
  • Pydantic AI · Optional
  • Exa · Optional
Orchestration
  • Pydantic AI · Optional

Evaluation contract

  • exact_entity_match · Weight 1
  • f1_score · Weight 1

Metrics

  • f1_score · Target 0.95

Benchmarks

  • FinAML-Graph-Bench
  • security

Reference cases

  • 1 public reference case

Workflow tasks

  1. Task 1

    Input: JSON graph adjacency list of transactions

    Output: JSON list of flagged high-risk entities

    Execution harness: Python

Evaluation tools

  • LlamaIndex · Knowledge Graph Framework
  • Pydantic AI · Agent Validation Framework

Recipe freeze marker

No recipe freeze time is recorded.

Frequently Asked Questions about COSMIC Intelligence Sharing Graph Agent with LlamaIndex and Pydantic AI