Research & Development

What We Explore

Advancing Technology through Research

Philosophy

Our Drive

Detect. Repair. Automate.

Developers spend countless hours finding and fixing software defects. As software grows more complex, so does the burden — and relying on individual expertise makes long-term maintenance increasingly difficult.

We research technologies that automate the detection and repair of software defects. Our goal goes beyond reducing the burden on developers. We want to make software more trustworthy for everyone who relies on it. As these technologies become more widely adopted, we hope developers can reclaim their time — and focus on creating what only they can create.

Fields

Core Expertise

Software Testing

We research techniques for systematically detecting and fixing software defects, including mutation testing and automated program repair. Our work also covers automated generation and transformation of test cases, and detection of non-deterministic defects in Web applications.

Web App Analysis

We research techniques for modeling, analyzing, and verifying the dynamic behavior specific to Web applications — covering both client-side asynchronous processing and dynamic UIs, and server-side MVC architectures — through a state machine-based approach.

AI & ML Engineering

We develop frameworks supporting testing and automated repair of deep learning models (DNNs), focusing on verifying that DNNs satisfy intended behavioral properties. Our work also includes automated model tuning using evolutionary computation.

Applied Data Science

We apply software engineering and data analytics techniques to scientific domains. Our work includes LLM-based automation of drug discovery data curation and the development of learning analytics infrastructure.

Publications

Selected Achievements

  • Conference Technical Briefing on Deep Neural Network Repair IEEE/ACM International Conference on Software Engineering (ICSE'24)
  • Journal DLiP-PPI library: An integrated chemical database of small-to-medium-sized molecules targeting protein-protein interactions Frontiers in Chemistry, 2023
  • Conference Validating Ajax Applications Using a Delay-Based Mutation Technique ACM/IEEE International Conference on Automated Software Engineering (ASE'14)
  • Conference Automated Verification of Pattern-Based Interaction Invariants in Ajax Applications IEEE/ACM International Conference on Automated Software Engineering (ASE'13)
View all publications & grants

OSS

Community Contributions

DNN Repair Framework

A framework for automatically tuning deep learning models to satisfy specific specifications using evolutionary computation. Developed and published as part of the eAI-QAML project.

View on GitHub

Dockerized Learning Analytics Platform

A Docker-based Learning Analytics platform for collecting, analyzing, and visualizing learning logs. Published by the Research Center for Open Science and Data Platform at the National Institute of Informatics.

View on GitHub

Collaboration

Advance Research Together

We welcome opportunities for joint research and technical validation with universities, research institutions, and industry partners.

Go to Contact Form