Skip to main content
Loading navigation
Back to challenges

Google ADK Federal Advisory Strategy Analyzer with Gemini

Analyzing public health advisory testimony and strategic plan documents demands multimodal document processing and compliance checking. Build a Google ADK agent powered by Gemini to parse multi-modal hearing transcripts and extract prioritized research funding allocations. Ensure complete source citation verification across 100 percent of extracted metrics.

Data ScienceHosted by Vera
Sign in free to start

What you are building

Build a multimodal public health intelligence agent using Google ADK and Gemini to extract, structure, and verify strategic research allocations from committee records.

How work is evaluated

Evaluation

Evaluates citation accuracy, funding priority extraction precision, and response generation quality.

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

citation_completeness

Checks if every extracted priority item includes a timestamp citation.

Binary check

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

Dimension 2

extraction_precision

Precision of extracted budget numbers against ground truth 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

  • Initialize Google Agent Development Kit (ADK) agent structure with Gemini 1.5 Pro

  • Process multimodal inputs (audio transcripts, PDF policy briefs, presentation charts)

  • Extract structured research grant priority metrics and budget target recommendations

  • Verify citation alignment against original transcript timestamps

Resources

Dataset notes

Transcripts and presentation slides from federal advisory committee meetings regarding autism research strategic plans.

Challenge setup

Evaluates citation accuracy, funding priority extraction precision, and response generation quality.

Challenge input

JSON containing transcript_text and document_image_urls

Google ADK

Required framework for building Vertex AI agent architectures.

Gemini

Provides multi-hour context length and multimodal ingestion.

Evaluated output

JSON with strategic_priorities array, target_funding_millions, and citations

Evaluation checks
  • Checks if every extracted priority item includes a timestamp citation.
  • Precision of extracted budget numbers against ground truth text. • target: 0.95 • range: 0-1
Configured evaluation evidence
  • Benchmark: PolicyDoc-Bench
  • Extraction Precision 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
  • Google · Required
  • Google Gemini · Optional
  • Google ADK · Optional

Evaluation contract

  • citation_completeness · Weight 1
  • extraction_precision · Weight 1

Metrics

  • extraction_precision · Target 0.95

Benchmarks

  • PolicyDoc-Bench
  • general

Reference cases

  • 1 public reference case

Workflow tasks

  1. Task 1

    Input: JSON containing transcript_text and document_image_urls

    Output: JSON with strategic_priorities array, target_funding_millions, and citations

    Execution harness: Python

Evaluation tools

  • Google ADK · Agent Development Framework
  • Gemini · Multimodal LLM Provider

Recipe freeze marker

No recipe freeze time is recorded.

Frequently Asked Questions about Google ADK Federal Advisory Strategy Analyzer with Gemini