Striveworks Secures Patent for Innovative Data Lineage Process
The new process grants greater transparency into machine learning workflows and allows for better data governance.
Originally published on January 4, 2024 at PR.com.
AUSTIN, TX (January 4, 2024) — Striveworks has secured a patent for a new data lineage process that represents a step change in transparency and auditability for the machine learning industry.
Striveworks’ patented process automatically tracks data and activities throughout machine learning workflows—saving time and money for data science teams and allowing full transparency over machine learning operations (MLOps) processes for management, regulators, and auditors. With this innovation, no software alterations or further actions from developers are needed to capture data lineage, even when external services and databases are involved.
“With machine learning workflows, you have steps, and you have input data. If you put those two things together, you may expect that you’d get the same answer every time. But if you don’t also record the additional services providing information into that workflow, along with the states they were in when those interactions happened, you may rerun that workflow and get different results,” says Matthew Griffin, the Striveworks software engineer who was awarded the patent.
“Without this process, developers would need to build a custom one and rigorously enforce compliance across their teams. This process ensures that happens as part of the normal workflow, so developers can focus on their real objectives.”
The Striveworks data lineage process is a significant improvement over the current standard in MLOps. Now, users can more easily observe activities, outputs, and interactions with external services that happen as part of their machine learning workflows. Equipped with this knowledge, users can make more informed decisions about whether to repeat workflows, make non-destructive changes to workflows, or revert to previous states when needed. This process also enables a critical function for modern MLOps: the ability to remediate divergent models to generate consistent, useful results that foster trust in the technology.
“Our mission at Striveworks is to make artificial intelligence and machine learning models safe, trustable, and seamless parts of business-critical operations,” says Striveworks CEO Jim Rebesco. “With our platform’s lineage system, our customers can be confident that the outputs from, and interactions with, production models are fully observable. This process fosters transparency and enables remediation of failing models, both of which are essential to our vision of making MLOps disappear.”
About Striveworks
Striveworks is a pioneer in responsible MLOps for national security and other highly regulated spaces. Striveworks’ MLOps platform enables organizations to deploy AI/ML models at scale while maintaining full audit and remediation capabilities. Founded in 2018, Striveworks was highlighted as an exemplar in the National Security Commission on Artificial Intelligence 2021 Final Report. In 2023, Striveworks was recognized on the Deloitte Technology Fast 500TM as one of North America’s fastest-growing companies in technology. For more information, visit www.striveworks.com.
###
