Skip to main content
Loading navigation
Back to challenges

Build Startup IPO Regulatory Auditor with LlamaIndex Advanced RAG

Analyzing complex IPO draft prospectuses (DRHP) requires high-precision semantic retrieval across hundreds of financial pages. Build a advanced RAG query engine using LlamaIndex to query regulatory filings and extract key financial risks with sub-second retrieval accuracy.

Business OperationsHosted by Vera
Sign in free to start

What you are building

Build an IPO prospectus regulatory auditing agent using LlamaIndex advanced RAG and hierarchical vector indexing.

How work is evaluated

Evaluation

Evaluates retrieval recall and factual correctness of risk analysis extracted from startup IPO filings.

Datasets

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

How submissions are scored

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

chunk_attribution_present

Ensures retrieved answer cites specific page node source chunk

Binary check

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

Dimension 2

faithfulness

Factual consistency of answer relative to source DRHP text • 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

  • Build hierarchical document index trees with LlamaIndex

  • Implement hybrid vector and keyword search retrievers for financial filings

  • Automate key risk factor extraction from SEBI DRHP filing prospectuses

Resources

Dataset notes

Text extracts and tables from SEBI draft red herring prospectuses (DRHP) filed in 2026.

Challenge setup

Evaluates retrieval recall and factual correctness of risk analysis extracted from startup IPO filings.

Challenge input

JSON containing company_name and target_metric string

LlamaIndex

Provides structured document indexing and multi-document RAG capabilities

Evaluated output

JSON containing answer, source_chunk_id, and faithfulness_score

Evaluation checks
  • Ensures retrieved answer cites specific page node source chunk
  • Factual consistency of answer relative to source DRHP text • target: 0.95 • range: 0-1
Configured evaluation evidence
  • Benchmark: DRHP-RAG-Bench
  • Faithfulness target: 0.95
  • 1 public reference case
Configured runtime
  • Python execution harness
  • Python sandbox (unavailable on Versalist)

Shows the first task, up to three tools, and up to three evidence items.

View technical recipe

Configured tools

Action Space
  • Llama Index · Required
  • Exa · Optional
Observation
  • Braintrust · Optional
Reward / Eval
  • Braintrust · Optional

Evaluation contract

  • chunk_attribution_present · Weight 1
  • faithfulness · Weight 1

Metrics

  • faithfulness · Target 0.95

Benchmarks

  • DRHP-RAG-Bench
  • general

Reference cases

  • 1 public reference case

Workflow tasks

  1. Task 1

    Input: JSON containing company_name and target_metric string

    Output: JSON containing answer, source_chunk_id, and faithfulness_score

    Execution harness: Python

Evaluation tools

  • LlamaIndex · Agent / RAG Framework

Recipe freeze marker

No recipe freeze time is recorded.

Frequently Asked Questions about Build Startup IPO Regulatory Auditor with LlamaIndex Advanced RAG