FAQ

Frequently Asked Questions

Everything you need to know about Panoptes and how it works.

What is Panoptes?

Panoptes is a computer vision platform that uses Vision Language Models (VLMs) to convert shop-floor video into structured, categorized insights about manufacturing events.

It combines:

  • Video from production lines
  • MES downtime and machine event data
  • AI-generated descriptions and classifications

The goal is to improve root cause analysis, downtime attribution, and operational decision-making.

What business problem is Panoptes solving?

Panoptes addresses the challenge of "undefined" downtime events, where production stops but the root cause is unclear.

  • A significant portion of downtime is not accurately categorized (>20%)
  • This limits OEE accuracy and slows root cause analysis
  • Panoptes uses AI to generate insights and categorizations to define these events
How does Panoptes work?

Panoptes follows an event-driven workflow:

  1. MES identifies a downtime event
  2. A corresponding video window is retrieved
  3. A VLM analyzes the video
  4. The system generates:
    • A description of what occurred
    • A categorized downtime reason
    • Structured output for analysis and reporting

Outputs are then made available through a UI aligned to and filterable by event timelines and machine location.

What is a Vision Language Model (VLM)?

A VLM is an AI model that can:

  • Analyze video or images
  • Understand visual scenes
  • Generate natural language descriptions

In Panoptes, VLMs are used to:

  • Describe machine behavior preceding a downtime event
  • Categorize events into informative reason codes
  • Provide clear and actionable insights for use in Root Cause analysis
Where does Panoptes run?

Panoptes is designed with an edge-first architecture.

  • Core processing runs on edge devices located in each plant
  • These devices host:
    • Video ingestion (e.g., camera feeds)
    • VLM inference
    • Local data storage and APIs

Each facility operates its own instance of Panoptes, enabling distributed deployment.

Why does Panoptes use edge devices instead of relying fully on the cloud?

The platform prioritizes edge deployment to:

  • Reduce network bandwidth usage by avoiding full video streaming
  • Lower cloud compute costs for VLM inference
  • Improve performance and latency
  • Keep processing within the plant network

Video is processed locally and only relevant data is transferred upstream.

Does Panoptes replace your existing MES or SCADA system?

No. Panoptes augments your existing systems by adding a visual intelligence layer. It integrates with your MES data to provide the visual context that reason codes and PLC fault data alone cannot capture.

What systems does Panoptes integrate with?

Panoptes integrates with several factory and enterprise systems:

  • Snowflake for MES data retrieval
  • Camera infrastructure for video capture
  • Edge compute infrastructure for processing
  • Enterprise identity systems for authentication and access control

Integration focuses on aligning machine data with video for synchronized analysis.

What are the primary use cases for Panoptes?

Current documented use cases include:

  • Multi-angle video replay of downtime events
  • Highlight clips tied to MES-triggered events
  • AI-generated descriptions of machine behavior
  • Event categorization aligned to downtime taxonomies
  • Image-based inspection that leverages a VLM instead of traditional training methodologies
How does Panoptes support root cause analysis?

Panoptes improves root cause analysis by:

  • Linking video directly to downtime events
  • Providing AI-generated descriptions of what happened
  • Enabling multi-camera replay for full context
  • Structuring previously unstructured observations

This reduces reliance on manual interpretation and operator memory.

What data does Panoptes use?

Panoptes uses a combination of:

  • Video footage from fixed cameras
  • MES downtime and fault events
  • Machine state and process data
  • Metadata linking cameras to production stations

Only event-relevant video segments are analyzed to optimize storage and compute usage.

How is Panoptes different from traditional computer vision?

Traditional computer vision focuses on:

  • Binary detection (good vs bad)
  • Pre-trained models for specific use cases

Panoptes extends this by:

  • Using VLMs for contextual understanding
  • Generating natural language explanations
  • Supporting flexible, prompt-driven scenarios
  • Aligning insights directly with MES-driven workflows
What is the scope of the Panoptes Project?

Panoptes is a test case that has two phases:

  • Proof of Concept (POC) — to validate feasibility of using a VLM to generate accurate and actionable categorizations and descriptions of machine behaviors
  • Proof of Value (POV) — to validate impact of Panoptes when running with live production data
What are the expected benefits of Panoptes?

The benefits include:

  • Reduction in undefined downtime events
  • Faster root cause analysis
  • Improved OEE reporting accuracy
  • Better operational decision-making
  • Increased production throughput
  • Scalable capability across plants and lines
How does Panoptes scale across plants?

Panoptes is designed for scalability through:

  • Edge-based deployments per facility
  • Standardized architecture and pipelines
  • Reusable data integration patterns
  • Modular AI workflows

Panoptes is scalable locally (additional lines at the same plant) and regionally (across different plants) once validated.

What are the main risks or challenges?

Current documented challenges include:

  • Data quality and MES alignment
  • Variability between production lines
  • VLM accuracy at scale
  • Integration with existing systems
How is security handled?

Security considerations include:

  • Authentication via enterprise identity systems
  • Role-based access control
  • Local processing within the plant network
  • Logging and monitoring for traceability
Will Panoptes replace MES or other factory systems?

No. Panoptes is designed to augment existing systems, not replace them.

  • MES remains the system of record for events
  • Panoptes adds visual context and AI interpretation
  • Outputs can be integrated back into existing workflows
What is the long-term vision for Panoptes?

The long-term vision is to establish a scalable, reusable AI capability that:

  • Enhances how downtime is analyzed
  • Enables broader computer vision use cases
  • Supports continuous improvement across manufacturing operations

Panoptes is meant to be a platform that can expand to multiple use cases and sites over time.

What capabilities are expected in future phases?

Future considerations include:

  • Expanded prompt and scenario management
  • Aggregated insights across time and facilities
  • Collaboration and feedback tools
  • Integration with additional enterprise systems (i.e. Maximo)

These are referenced as future enhancements and not part of the current project scope.

Who should use Panoptes?

Primary users include:

  • Plant operations teams
  • Maintenance and engineering teams
  • Operational Excellence (OEX) leaders
  • Analysts responsible for OEE and downtime performance