Skip to main content
Back to challenges
Challenge

Orchestrating Multi-Agent Building Conversion Engineering with CrewAI and Patronus AI

The Pfizer building scare highlighted that office-to-residential conversions are primarily complex engineering challenges rather than simple architectural updates. This challenge requires developers to build a multi-agent system using CrewAI to conduct structural and regulatory due diligence for commercial real estate conversions. You will orchestrate three distinct agents: a Structural Systems Analyst, a Regulatory Compliance Officer, and a Financial Underwriter. These agents must collaborate to identify 'red flag' structural risks (e.g., floor plate depth, plumbing stack capacity) from provided building specifications. To ensure the reliability of the agents' technical advice, you will integrate Patronus AI for automated evaluation. The system must use Patronus AI's 'Lynx' model or similar validators to detect hallucinations in structural code citations and verify that the proposed engineering solutions align with the International Building Code (IBC) standards for residential conversions. The goal is a robust pipeline that flags technical risks before they become financial liabilities.

Business OperationsHosted 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 multi-agent workflow to evaluate office-to-residential conversion feasibility using CrewAI, validated by Patronus AI for engineering accuracy.

Delivery guide

How work is evaluated

Evaluation

Evaluation focuses on the agent's ability to identify specific engineering risks and the validation of those claims via Patronus AI.

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

Patronus Accuracy Threshold

The output must pass the Patronus AI Lynx hallucination check.

Binary check

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

Dimension 2

Risk Recall

Percentage of pre-identified structural risks detected by the agents. • target: 0.9 • 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

  • Orchestrate specialized role-playing agents using CrewAI to solve multi-domain engineering problems

  • Implement inter-agent communication and task delegation for complex real estate due diligence

  • Integrate Patronus AI to validate the accuracy of LLM-generated technical advice and code compliance

  • Analyze building floor plates and MEP (Mechanical, Electrical, Plumbing) specs via LLM-driven agents

Resources and assets

Reference links and supporting material

Dataset notes

Synthetic building specification sheets including floor plans, MEP layouts, and current office occupancy data based on common 1980s commercial designs.

Challenge setup

Evaluation focuses on the agent's ability to identify specific engineering risks and the validation of those claims via Patronus AI.

Challenge input

JSON containing building floor plates, HVAC specs, and plumbing capacity

CrewAI

Allows specialized agents to collaborate on the conversion problem.

Patronus AI

Used to ensure engineering advice is grounded in fact.

Evaluated output

JSON report with identified risks, IBC citations, and feasibility score

Evaluation checks
  • The output must pass the Patronus AI Lynx hallucination check.
  • Percentage of pre-identified structural risks detected by the agents. • target: 0.9 • range: 0-1
Configured evaluation evidence
  • Benchmark: CRE-Agent-Eval
  • Risk Recall target: 0.9
  • 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
  • CrewAI · Optional
  • Synthflow · Optional
Reward / Eval
  • Patronus AI · Required
Safety / Guardrails
  • Patronus AI · Required

Evaluation contract

  • Patronus Accuracy Threshold · Weight 1
  • Risk Recall · Weight 1

Metrics

  • Risk Recall · Target 0.9

Benchmarks

  • CRE-Agent-Eval
  • general

Reference cases

  • 1 public reference case

Workflow tasks

  1. Task 1

    Input: JSON containing building floor plates, HVAC specs, and plumbing capacity

    Output: JSON report with identified risks, IBC citations, and feasibility score

    Execution harness: Python

Evaluation tools

  • CrewAI · Orchestration Framework
  • Patronus AI · Validation

Recipe freeze marker

No recipe freeze time is recorded.

Frequently Asked Questions about Orchestrating Multi-Agent Building Conversion Engineering with CrewAI and Patronus AI