I am pleased to announce that I have received a three-year standard research grant from the National Science Foundation (NSF) Software & Hardware Foundations (SHF) program as principal investigator (PI) for a project entitled “Knowledge, Methodologies, and Tool-support for Combating Technical Debt in Machine Learning Systems.” The total grant amount is ~$600K.

A Machine Learning (ML) system is a software system where at least one component of the system involves learning, i.e., it was not programmed by code but rather data. In treating data as “code” and consisting of many subsystems that support learning, unfortunately, ML systems exhibit not only problems of typical software but also those unique to the paradigm. A central issue in ML systems is technical debt, where software developers trade off short-term gains in bug fixes and new software features for long-term quality. Technical debt makes modifying ML systems difficult and error-prone, which can negatively impact its effectiveness.

The project will facilitate the long-term reliability and evolvability of ML systems by investigating how:

1. Data and multiple model interactions influence technical debt.
2. Object-Oriented Programming (OOP) can be used to reduce technical debt by encapsulating model code representing subtle ML algorithm differences.
3. Data can be abstracted and refactored (semantics-preserving, source-to-source program transformation) alongside code to combat technical debt.

More information may be found on NSF’s website; stay tuned for more details and funded research opportunities!