BIM managers and AEC professionals spend excessive time manually parsing dense model data to verify compliance and identify life-safety risks. There is a massive opportunity to leverage AI to shift this workflow from manual checking to automated intelligence orchestration.

By building a seamless integration with Autodesk’s existing environment (BIM 360 / ACC), Trace can drastically reduce manual review time, prevent costly downstream oversights during construction, and empower managers to focus on critical decision-making rather than data hunting.

• Beyond the Chatbot

  AEC professionals do not need another chat interface.
  They need structured signal clarity.

• Agentic Architecture

  Instead of one monolithic AI, Trace operates as:
  Orchestrator of Agents. In V0, the system begins with three core agents Each agent processes a different intelligence layer.

  The Orchestrator synthesizes outputs into a unified Risk Report.

• Trust & Explainability

  In regulated AEC environments, AI must be accountable.

  Given my research background in trust within virtual systems, Trace is intentionally designed to be explainable and auditable.

  The system emphasizes:

  • Traceable reasoning chains (how a signal was generated)

  • Transparent categorization (severity × criticality logic visible)

  • Confidence signals (cluster amplification and weighting logic exposed)

  • Clear remediation pathways (structured issues pushed to ACC)

Merges intelligence from A + B + C into a single Risk Report.

For each signal from Agent A:

 1. Look up criticality from Agent B for the same element

 2. Compute weighted_score = severity_weight × criticality_weight

 3. Check if signal is part of a cluster from Agent C

 4. If clustered: weighted_score × amplification_factor

5. Sort all findings by weighted_score descending

6. Group by zone/level

7. Compute overall risk = f(weighted_scores, cluster_count, failure_rates)

8. Generate narrative summary

1. User selects ACC project + model

2. Ingest Layer pulls element metadata via API

3. Records normalized → stored

4. Orchestrator dispatches to Agent A + Agent B (parallel)

5. Agent A returns compliance signals

6. Agent B returns criticality weights

7. Orchestrator dispatches to Agent C (sequential — needs A+B)

8. Agent C returns risk clusters

9. Orchestrator synthesizes → Risk Report

10. Dashboard renders report

11. User [ in V0: BIM manager] approves → issues pushed to ACC

1. Frictionless Authentication: The user signs in seamlessly using their Autodesk credentials, securely authorizing the application to access their ACC Document Management hubs.

2. Native Asset Selection: A clean picker interface allows users to naturally navigate their ACC structure to select their project and target Revit model.

3. Orchestrated Analysis: Upon initiating the analysis, the app provides real-time progress while asynchronously polling Autodesk APS for translations and routing the payload through the AI pipeline.

4. Actionable Review: The user is immediately presented with a highly visual, graded risk report highlighting critical clusters and hotspot levels, allowing BIM managers to target non-compliant design elements before they reach the job site

• Seamless Autodesk ACC Integration

• Automated Ingestion & Translation

• Multi-Agent Risk Pipeline

• Executive Risk Dashboard

Developed a functional, production-ready MVP that successfully authenticates with Autodesk's ecosystem, ingests real BIM data via the Model Derivative API, and translates raw architectural elements into an actionable, AI-driven risk report.

The platform serves as a powerful proof-of-concept demonstrating how targeted intelligence orchestration and modern UX can sit on top of Autodesk Platform Services to solve complex, high-value AEC workflows.