Making the World Safer for Robots and Humans

Distribution-free Evaluation and Error Correction for AI/ML

- What We Do -

Click here to see our entry to the
Stanford HAI 2022 AI Audit Challenge.


At Data Engines, we offer ground truth inference as a service.  Unleashing the power of unlabeled data,  GroundSeer™ leapfrogs the conventional "wisdom of the crowd" approach to provide better insight than simple majority rule, allowing us to auto-focus AI/ML decisions to the highest precision.

How does GroundSeer™ work?

Our patented technology uses algebraic statistics and plain linear algebra to discover the hidden pattern underneath the decisions of the crowd. This pattern can be used for:

  • Best-of-breed classifier selection
  • Error correction in near real-time for unlabeled data streams
  • Detection of malfunctioning classifiers or regressors

Ensure your data pipeline has its highest precision possible.

Questions? Contact us.

data_picture

- Applications -

GroundSeer™ has a variety of possible applications in the arenas of bioinformatics, national security, medical imaging, online advertising/marketing, and data retrieval.

It is particularly suited to artificial intelligence.

Illumination_Hand-BW
  • Self-driving cars can automatically detect which algorithms or sensors are failing or underperforming.

  • Robotic machines can perform self-assessments on sensors so as to discount bad or failing ones.

  • Security classifiers can detect “intrusion” signals to ensure security software operates using multiple algorithms simultaneously in a whole-is-greater-than-the-sum-of-its-parts scanning operation. Faulty cameras, for instance, could be detected instantly.

  • Image analysis and bioinformatics classifiers can recognize alterations in genetic sequences to enable a decision of whether cells are likely to be cancer or not. A similar assessment could be done using information extracted from images of cells or tissues.

  • Online, the degree of correlation among classifiers making labeling decisions can be continuously monitored.

  • Large datasets can be turned into training data with precise knowledge of the quality of the labels.

- Contact Us -

Let us talk to you about your data. We can help you reach your objectives through our customizable assessment and measurement techniques.

Please enter your name.
Please enter a message.

- White Papers and Other Information -

GroundSeer™

Measuring Performance Using the Performance Tracker

- Executive Biographies -