Low-rank imputation engine for AutoML pipeline selection
This technology is a recommendation engine that figures out which AI processing pipeline will work best for a given dataset—without having to actually run every possible combination to test it. It does this by filling in the gaps of a performance matrix (think: a spreadsheet where rows are pipelines and columns are datasets, but many cells are blank) using a math technique called local low-rank matrix completion. By predicting the missing performance scores from the patterns in nearby data, it can rank which pipeline is likely to perform best on a new dataset before any expensive compute is spent. The result is a practical AutoML-style configurator that learns from past experiments to make smart recommendations for new contexts.
What you could build
An AutoML or MLOps platform feature that automatically recommends the best AI pipeline configuration for enterprise datasets, reducing experimentation cost and time-to-deployment; target buyers would be data science platform vendors (e.g., DataRobot, H2O.ai competitors) and large enterprises running repeated ML workloads across diverse data contexts.
Who in Virginia should care
Northern Virginia's dense concentration of federal AI contractors, cloud providers (AWS, Microsoft, Google), and defense analytics firms would be natural evaluators of pipeline optimization tooling for large-scale government data workloads.
Readiness: Concept
Concept — described but not yet demonstrated. Lab validated — supported by experimental results in the patent. Prototype likely — the text describes a built, working embodiment.
Readiness is inferred from the patent text, not from a lab visit.
The record
- Inventors
- Ran Jin, Xiaoyu CHEN
- Filed
- Patent pending — filed October 25, 2024
- Status
- Application
- Publication number
- US20250299073A1
Ready to talk?
Virginia Tech Intellectual Properties handles licensing for this technology.
Prosim summaries are generated from public patent text and are not legal advice.