Skip to main content
Back to challenges
Challenge

Clinical Continuity Agent: Syphilis Treatment Protocol & Drug Shortage Orchestrator

Public health infrastructure is currently struggling with a resurgence of syphilis compounded by severe shortages of Bicillin L-A and restrictive federal research constraints. This challenge tasks you with building an intelligent clinical decision support agent using LlamaIndex that dynamically adjusts treatment recommendations based on real-time pharmacy inventory levels and patient-specific contraindications. You will implement advanced RAG techniques to parse complex CDC clinical guidelines and merge them with unstructured pharmacy availability reports. Your agent must navigate the 'federal constraints' mentioned in STAT news by prioritizing protocols that comply with current NIH and FDA regulatory frameworks while identifying alternative pharmacotherapy routes. Using Aider as your AI pair-programmer, you will develop a modular system that can ingest FHIR-formatted patient records and output structured treatment plans with full provenance links to the source literature. The solution must demonstrate robust handling of 'drug-out' scenarios where primary treatments are unavailable, suggesting secondary or tertiary alternatives based on efficacy and patient allergy profiles.

Frontier Science & ResearchHosted 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 LlamaIndex-powered clinical agent that orchestrates syphilis treatment protocols by cross-referencing CDC guidelines with dynamic drug inventory data.

Delivery guide

How work is evaluated

Evaluation

Evaluation focuses on clinical accuracy, adherence to the shortage logic, and data structure validity.

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

Safety Check: Pregnancy

Ensure Doxycycline is NEVER suggested for pregnant patients even during shortages.

Binary check

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

Dimension 2

Citation Accuracy

Percentage of recommendations containing valid CDC/NIH links. • target: 1 • 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

  • Master LlamaIndex 'Query Pipelines' for complex multi-stage clinical reasoning

  • Implement Aider-driven development workflows for rapid healthcare microservice prototyping

  • Design a semantic retrieval system for high-stakes medical protocols with 100% citation accuracy

  • Build a logic-driven fallback mechanism for pharmaceutical shortages using knowledge graphs

  • Integrate FHIR (Fast Healthcare Interoperability Resources) data models into LLM context windows

Resources and assets

Reference links and supporting material

Dataset notes

Synthetic dataset containing 50 patient FHIR records, 10 CDC Syphilis protocol PDFs, and a live-updating CSV of drug inventory levels across 5 San Diego clinics.

Challenge setup

Evaluation focuses on clinical accuracy, adherence to the shortage logic, and data structure validity.

Challenge input

JSON containing patient profile and inventory status

LlamaIndex

Best-in-class for RAG and data-connected agents.

Aider

Speeds up the creation of complex clinical parsing logic.

FastAPI

Exposes the agent as a clinical microservice.

Evaluated output

Structured Treatment Plan JSON

Evaluation checks
  • Ensure Doxycycline is NEVER suggested for pregnant patients even during shortages.
  • Percentage of recommendations containing valid CDC/NIH links. • target: 1 • range: 0-1
Configured evaluation evidence
  • Benchmark: ClinicalQA / MedQA
  • Citation Accuracy target: 1
  • 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
  • Aider · Required
  • Lingo · Optional
  • Llama Index · Optional
Observation
  • Aider · Required

Evaluation contract

  • Safety Check: Pregnancy · Weight 1
  • Citation Accuracy · Weight 1

Metrics

  • Citation Accuracy · Target 1

Benchmarks

  • ClinicalQA / MedQA
  • biotech

Reference cases

  • 1 public reference case

Workflow tasks

  1. Task 1

    Input: JSON containing patient profile and inventory status

    Output: Structured Treatment Plan JSON

    Execution harness: Python

Evaluation tools

  • LlamaIndex · Orchestration Framework
  • Aider · AI Engineering
  • FastAPI · API Layer

Recipe freeze marker

No recipe freeze time is recorded.

Frequently Asked Questions about Clinical Continuity Agent: Syphilis Treatment Protocol & Drug Shortage Orchestrator