The FUSE Lab centers its research on one core theme: opportunities and challenges of leveraging AI in software engineering at the scale of tens of thousands of engineers.
To drive these advancements, the lab leverages key technical AI components, focusing primarily on:
With a commitment to explainable AI systems.
The Future of Software Engineering Laboratory (FUSE) is dedicated to bridging the gap between cutting-edge AI research and practical software engineering at scale.
The lab’s mission and vision are to rethink the future of software development by building the fundamental knowledge base, technological tools, and human capacities necessary to transition to a new generation of software engineering workflows, embracing AI as a force multiplier while addressing the key challenges and risks it introduces.
The impact of the lab lies in accelerating developer productivity, code quality, and system reliability. This is achieved by validating breakthroughs directly in real-world environments, advancing open science to empower developers worldwide, and embedding safe, explainable, and human-centered AI frameworks to improve developer wellbeing and trust across the software ecosystem.
Automated code refactoring – Explores how to use Large Language Models (LLMs) to automatically refactor large-scale codebases. The goal is to reduce technical debt and complexity while increasing overall code quality.
Automated test generation – Focuses on improving software testing, this project investigates coverage effectiveness and test prioritization. It aims to define what “good coverage” means and leverages Meta’s models to improve industry standards.
Code review – Leverages AI to enhance the code review process. It tackles challenges such as reviewer recommendations, improving the quality of comments, automated fixing of identified issues, and reducing review latency.
Engineering productivity metrics – Aims to define and capture productivity standards for software and Machine Learning Engineers (MLE). It focuses on product metrics (e.g., code complexity), process metrics (e.g., review latency), and people metrics (e.g., expertise levels).
The dual-use dilemma of LLMs4Code – Investigates the security implications of using LLMs in software engineering. It explores how these models can improve software security (e.g., binary decompilation) while also addressing risks like malicious use, data leakage, and intellectual property conflicts.
Delft University of Technology (TU Delft) is a leading academic institution where the lab is based, specifically within the Software Engineering Research Group.
Meta is a global technology conglomerate known for social technologies that help people connect, find communities, and grow businesses. Within the FUSE Lab, the specific focus is on their Developer Infrastructure (DevInfra) team.
