I-Corps: Translation potential of next generation machine learning pipelines for numeric and tabular data

Award ID: 2534217

This I-Corps project is based on the development of an artificial intelligence (AI) framework for analyzing numeric and tabular data. Numeric data plays a crucial role in business operations and decision-making, whether financial data, supply chain data, consumer metrics, or patient health data. Currently, data professionals want easy-to-use tools for numeric and tabular data that have a simple workflow and integrate seamlessly into their existing processes, while providing the flexibility and power required for complex data analysis. There is a growing demand for AI-assisted advanced analytics and decision-making, however, advancements in AI have made limited progress with numeric and tabular data. This technology is designed to improve the efficiency and productivity of data professionals. The goal is to improve access to advanced analytical tools and bolster public engagement with AI while helping ease the impact of skilled labor shortages. The technology is industry-agnostic and may help to create new products, generate jobs, and enhance economic activity. The data analysis framework also may address systemic inefficiencies and improve practices related to analyzing numeric and tabular data. This I-Corps project utilizes experiential learning coupled with a first-hand investigation of the industry ecosystem to assess the translation potential of using an artificial intelligence (AI) technology for numeric data analysis. The rapid advancements in large language models (LLMs) have sparked enthusiasm for AI, but extending these advances to numeric and tabular data has had comparatively limited progress. The technology uses a new class of machine learning (ML) pipelines based on a novel hierarchical neuro-symbolic architecture. Classical as well as deep neural models require iterative rework in manual and ad hoc tuning of their features or model architectures. These challenges are addressed using a hierarchical architecture that is interpretable, can use domain knowledge, and is simplified. This solution may provide data professionals with an easy-to-use tool for numeric and tabular data while giving them the flexibility and power required for complex data analysis. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

NSF Program Director: Ruth Shuman