This project is under NDA. Below is a summary. Reach out for process and final designs.

What

Our customers install networked cameras across factory floors, construction sites, and public areas to monitor:

• Equipment inspection

• Hazardous zone identification

• Object counting

• Predictive maintenance

• Analog instrument reading

• Crowd analytics

They need a single interface to watch live feeds, review recordings, interact with video (pause, seek, screenshot), filter by use-case or camera group, and view alerts and analytics—all without switching tools.

Why

Existing video systems force operators to juggle separate players, spreadsheets, and email alerts. This creates friction:

• Fragmented UI: Live streams, recordings, and analytics live in different apps

• Manual workflows: Teams download clips, then run analytics in Excel—slow and error-prone

• Poor filtering: Hard to focus on specific cameras, floors, or items (e.g., helmet compliance)

• Alert fatigue: Notifications aren’t tied back to video evidence, so context is lost

By unifying video control, filtering, alerts, and reporting, we reduce response time and improve situational awareness.

User and Business Goals

Users

• Operators & Safety Managers need:

  • One-click expand of any feed (live or recorded)

  • Seamless pause, rewind, and screenshot

  • Dropdown filters to focus on helmets, PPE, or specific floor views

  • Real-time alerts linked to video evidence
    
    

• Analysts & Engineers need:

  • Unified dashboard for video-driven KPIs (counts, readings, compliance rates)

  • Ability to bookmark clips and export reports

Business:

• Reduce incident response time by surfacing evidence instantly

• Automate reporting to cut down manual data aggregation

• Scale the platform to support hundreds of cameras and multiple use cases without new development
  
  

How

We followed an incremental, UX-driven process:

1. Discovery & Research

   • Interviewed site supervisors, maintenance teams, and safety officers

   • Mapped existing toolchains and pain points

   • Analyzed competitor flows (e.g. viso.ai) for best practices
     
     

2. Initial Prototypes

   • Used Cursor-generated React components to build a working multi-camera grid and video player

   • Tested live/recorded toggle, expand on click, and playback controls with operators
     
     

3. Iterative Design & Validation

   • Introduced dropdown filters for use case (helmet, hazard zone) and camera groups (floor, area)

   • Designed alert center that links each notification to a video snippet

   • Built analytics dashboard wireframes showing object counts over time and analog readings
     
     

4. Collaboration with Developers

   • Defined API contracts for video feeds, metadata, and alert events

   • Integrated Cursor-generated UI with backend services

   • Conducted usability sessions on staging build to refine microcopy and toolbar access

Potential Impact

Once live, we expect to see:

• 30–50% faster incident investigation, as operators click directly from alerts to video

• 20% reduction in manual reporting hours with built-in analytics exports

• Higher compliance through real-time PPE monitoring and alerts

Continuous Improvement

Future epics planned post-launch:

• Advanced AI Insights: Automated root-cause suggestions (e.g., vibration spikes → maintenance ticket)

• Mobile App: Critical-alerts only view for field supervisors

• Custom Reporting Templates: Drag-and-drop report builder with video embed support

• User-Defined Workflows: Save and share common filter sets and camera layouts

This case study highlights how we turned complex, fragmented video monitoring into a cohesive, user-centric AI vision platform ready to scale.