CalypsoAI Solutions for the Public Sector

VESPR Validate

VESPR Validate is an independent model validation platform with built-in, industry leading model robustness testing. VESPR Validate tests model performance against evasion and privacy attacks. It also stress tests models against real-world conditions (e.g. image quality, moving subjects, adverse weather conditions, image compression). It provides human oversight for decision making. Standardized model reports are generated for stakeholders or external reviewers. The report encapsulates model performance, the context the model is used in and model validation decision rationale.

  • Computer Vision AI

    Computer Vision AI

    As computer vision increasingly takes over critical business functions, there is an urgent need for independent model validation. The global computer vision market is valued at over US $12bn and is expected to reach US $20bn by 2028[1]. The market encompasses activities across all industries, from manufacturing and product development to corporate finance and human resources. Significantly, Quality Assurance and Assessment is the largest segment of this market accounting for 28% of global revenue in 2020.

    Within an AI governance framework, there is a need for independent model validation to assess model risk in the context of the specific business operation where the model will be deployed. Key model validation metrics are model performance, security, resilience and privacy. The need for independent model validation comes from the experience of industries such as banking that have been relying on models for over 20 years. Regulators across jurisdictions in these industries have mandated independent model review (e.g. Solvency II, TRIM, SR 11-7) largely in response to the events that led to the global financial crisis.

    VESPR Validate's suite of over 15 tests ensure AI models are independently validated. With an easy-to-use interface that facilitates cross-organizational collaboration, VESPR Validate provides a repeatable and measurable process that establishes model governance. As a result, users can confidently deploy their AI models into production.

    [1] Grand View Research. 2021. “Computer Vision Market Size, Share & Trends Analysis Report By Component, By Product Type, By Application (Quality Assurance & Inspection, 3D Visualization & Interactive 3D Modeling), By Vertical, By Region, And Segment Forecasts, 2021 - 2028.” Grand View Research.

  • Full-Motion Video (FMV)

    CalypsoAI's Full-Motion Video capability enhances model governance and validation with the following features:

    • PyTorch/ TensorFlow model testing
    • Testing models against data corruptions and physics-based weather simulations
    • Clear metrics on model robustness for real-world deployment
    • Intuitive workflow for non-data scientists to benchmark and compare models