Documentation Build on the Opus API Blog Insights, hackathons, and updates
Research Papers from the Opus team Trust Center Security, privacy, and compliance Careers Join the team building Opus

Resources

Sign up
← All blog posts
Blog Product · Community

Building an Opus workflow to monitor infrastructure damage

3 min read

The Opus NYU Hackathon saw students build workflows targeted at solving real-world problems, and the results were impressive.

The first place winners, Team FixMyRoad, built a centralized urban pothole and transport infrastructure damage reporting system. The team members told us they come from countries where potholes are a major problem, and they showcased how Opus can be deployed to address challenges with the physical infrastructure that underpins our daily lives.

See how they built it here:

Pothole Reporting System

The Pothole Reporting System is an intelligent automation pipeline designed to streamline the identification and repair of road infrastructure issues. By integrating AI validation with structured data management, the system ensures that maintenance teams can prioritize high-impact or emergency repairs efficiently.

The system the team built includes the following features:

  • AI Validation: A “Custom Agent” determines if a submitted report is a valid pothole.
  • Smart De-duplication: Identifies if a pothole has already been reported; duplicate reports increase the priority of the issue.
  • Emergency Escalation: Automatically determines the severity of the pothole and flags emergency cases for immediate human review.
  • Structured Reporting: Generates a “Final Issue Record” containing summaries, location data, and severity levels.
  • Multi-Channel Output: Syncs with Google Sheets, sends email notifications, and exports data for government authorities.

System Workflow

The workflow, on Opus, follows a linear logic path with several conditional branches:

  1. Workflow Input: Receives raw data from a citizen-facing front-end.
  2. Custom AI Agent: Processes the input to verify the nature of the report.
  3. Validation Check: If the report is a valid pothole, it proceeds to data import.
  4. De-duplication: Checks for existing entries in the database.
  5. Emergency Determination: Uses AI to assess risk factors (e.g., high-traffic areas, depth of the pothole).
  6. Human-in-the-Loop: Emergency cases are routed for manual confirmation before final dispatch.

Logic & Decision Nodes

  • Is Valid Pothole? (Conditional) – Filters irrelevant or non-maintenance reports.
  • Duplicate? (Database Query) – Raises priority when multiple citizens report the same issue.
  • Emergency Determiner (AI Classifier) – Flags high-risk issues needing a 24-hour response.
  • Confirm Emergency (Manual Trigger) – Adds human verification before dispatch.

Standards & Outputs

The system is designed to provide a structured “Final Issue Record” that maps directly to municipal maintenance databases. This record includes:

  • Issue Summary: A concise description of the hazard.
  • Location Metadata: Precise coordinates and site names.
  • Confidence Score: AI-generated metric of report accuracy.
  • Risk Level: Categorized based on road type and pothole size.

Prerequisites

  • Access to the workflow automation platform.
  • API keys for the NLP Custom Agent.
  • Google Sheets integration for data storage.

Installation

  1. Import the workflow JSON file into your automation environment.
  2. Connect your email service provider.
  3. Configure the Google Sheets “Export Data” node with your specific sheet ID.

Watch the workflow in action:

Built for regulated work.
Ready when you are.

Get Started Talk to Sales