Skip to main content
Back to challenges
Challenge

Construct an Automated Document Factory for Bank Ingestion using LlamaIndex

Financial document factories require resilient extraction from unstructured invoices and identity forms. Build a LlamaIndex RAG pipeline hosted on Replicate to parse multi-lingual banking forms and trigger automated ledger entries with 98% field-level accuracy.

CybersecurityHosted 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 an Automated Document Factory pipeline with LlamaIndex and Replicate models to parse unstructured banking documents into verified ledger transactions.

Delivery guide

How work is evaluated

Evaluation

Measures field-level extraction precision across multi-page bank invoices.

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

IBAN Validation Test

Extracted IBAN must match ISO 13616 standard.

Binary check

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

Dimension 2

Field Extraction Accuracy

Percentage of perfectly extracted document fields. • target: 0.98 • range: 0.95-1

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

  • Design a LlamaIndex data ingestion pipeline for unstructured financial PDFs and images.

  • Deploy open-source vision-language models on Replicate for high-precision OCR extraction.

  • Validate extracted numerical schemas against double-entry accounting constraints.

  • Implement automated error routing when confidence scores drop below defined thresholds.

Resources and assets

Reference links and supporting material

Dataset notes

Multi-lingual bank invoice and remittance document dataset with annotated ground truth.

Challenge setup

Measures field-level extraction precision across multi-page bank invoices.

Challenge input

String URL or raw bytes of bank document payload

LlamaIndex

Manages data connectors and extraction pipelines.

Replicate

Executes open-source OCR and vision models in cloud sandboxes.

Evaluated output

JSON containing total_amount, currency, seller_iban, confidence_score

Evaluation checks
  • Extracted IBAN must match ISO 13616 standard.
  • Percentage of perfectly extracted document fields. • target: 0.98 • range: 0.95-1
Configured evaluation evidence
  • Benchmark: DocFactoryBench
  • Field Extraction Accuracy target: 0.98
  • 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
  • LlamaIndex · Optional
  • Llama 3.2 · Optional

Evaluation contract

  • IBAN Validation Test · Weight 1
  • Field Extraction Accuracy · Weight 1

Metrics

  • Field Extraction Accuracy · Target 0.98

Benchmarks

  • DocFactoryBench
  • security

Reference cases

  • 1 public reference case

Workflow tasks

  1. Task 1

    Input: String URL or raw bytes of bank document payload

    Output: JSON containing total_amount, currency, seller_iban, confidence_score

    Execution harness: Python

Evaluation tools

  • LlamaIndex · Agent Framework
  • Replicate · Inference & Model Runtime

Recipe freeze marker

No recipe freeze time is recorded.

Frequently Asked Questions about Construct an Automated Document Factory for Bank Ingestion using LlamaIndex