// MODEL OPTIMIZATION AND PROMPT SYNTAX TERM

Model Monitoring

Model monitoring involves continuously watching a deployed model's performance and behavior to ensure it's working correctly.

TECHNICAL DEFINITION

Model monitoring is the continuous observation and analysis of a deployed machine learning model's performance, data inputs, and outputs in a production environment to detect issues like drift, bias, or performance degradation.

BACKGROUND

Prompt injection is a cybersecurity exploit and an attack vector in which innocuous-looking inputs are designed to cause unintended behavior in machine learning models, particularly large language models (LLMs). The attack takes advantage of the model's inability to distinguish between developer-defined prompts and user inputs to bypass safeguards and influence model behaviour. While LLMs are designed to follow trusted instructions, they can be manipulated into carrying out unintended responses through carefully crafted inputs.

READ MORE ON WIKIPEDIA

SYNONYMS & ALIASES

  • ML model observability
  • model health monitoring
  • production model monitoring

USAGE NOTE

Effective model monitoring is critical for maintaining the reliability and fairness of AI systems.

DEVELOPERS

Organizations developing technology related to Model Monitoring.

  • Arize AI

    Provides an AI observability platform designed to help teams monitor, explain, and troubleshoot ML models in production, detecting issues like data drift, performance degradation, and bias.

  • WhyLabs

    Offers an AI observability platform that helps teams monitor data pipelines and AI models for data quality, drift, and performance issues, ensuring model integrity and reliability.

  • Fiddler AI

    Provides an AI Observability Platform that helps enterprises monitor, explain, and analyze their AI models in production, addressing performance, bias, and data drift.

  • Seldon

    Offers MLOps tools, including Seldon Deploy, which provides robust model monitoring capabilities for drift detection, performance metrics, and explainability for deployed machine learning models.

  • Datadog

    Offers comprehensive observability for AI applications, including specific capabilities for monitoring machine learning models in production, tracking performance metrics, and detecting anomalies.

  • Superwise

    An AI observability platform built specifically for monitoring ML models in production, enabling teams to detect issues, debug faster, and ensure model health and performance.

  • Google Cloud (Vertex AI Model Monitoring)

    Google Cloud's Vertex AI platform includes dedicated model monitoring services that help detect drift, bias, and performance issues in deployed machine learning models, offering automated alerts and visualization.

  • IBM

    Through its Watson Studio and Cloud Pak for Data offerings, IBM provides AI governance and MLOps tools that include robust model monitoring capabilities for fairness, explainability, and performance of AI models.

RELATED TERMS IN MLOPS & DEPLOYMENT