Advancing AI-Driven Software Engineering Through Transparent and Open-Source Innovation
FUSE Lab (Future of Software Engineering Laboratory) established at the Delft University of Technology is officially the 58th ICAI lab, and is dedicated to rethinking the software development lifecycle through the lens of artificial intelligence. The lab addresses the serious challenge of leveraging AI within software engineering environments that operate at large scales. FUSE is part of the Software Engineering Research Group (SERG) and sponsored by Meta’s DevInfra team. This collaboration grew from a mutual interest in the opportunities of AI-driven engineering at large scales, ensuring that research outputs are validated in both industrial and open-source contexts. Most importantly, all the lab results will be openly available through open source prototypes and open access publications.
The strategic direction and operational success of the lab lay in the hands of an experienced team of directors and managers with a diverse backgrounds. The board team is formed of Arie van Deursen, Andy Zaidman, Nachi Nagappan, Carolin Brandt, and Venkatesh Chandrasekar. Together they ensure that the lab achieves its overall mission and goals, overseeing the collaborative work of their PhDs that bridge the gap between theoretical academic research and practical industrial application.
The overall mission of the lab is to rethink the future of software development to adapt to a rapidly and constantly changing technological environment. This is grounded in the principle that all advancements must be informed by sound theories that are rooted in data and demonstrated through working prototypes. By embracing AI as a transformative opportunity while addressing the unique challenges that can emerge from implementation, the lab seeks to empower the engineering community through a firm commitment to open science and open-source software.
There are five core research tracks as part of the scientific work of the lab, which are designed to improve the developer experience:
1. Automated Code Refactoring: Utilizing Large Language Models (LLMs) to automatically reduce technical debt and architectural complexity.
2. Automated Test Generation: Investing test prioritization and coverage effectiveness to improve software reliability.
3. Code Review: Advancing review efficacy through AI-driven reviewer recommendations and automated comment fixing.
4. Engineering Productivity Metrics: Capturing metrics across product, process, and people, including a focus on Machine Learning Engineers.
5. The Dual-Use Dilemma of LLMs4Code: Investing how LLMs can enhance software security while addressing novel vulnerabilities and data extraction risks.
More in-depth information regarding the lab’s mission, research projects, partners, and team, can be found on the FUSE lab website, and the ICAI website.
