Tulane University · School of Science & Engineering

Toward more dependable software and AI

The Laboratory for Software Design works on programming languages and software engineering. Much of our current work studies how to make AI-enabled systems more reliable, using ideas from modular reasoning.

Explore our research Read recent papers

What we work on

Most of our work comes back to modular reasoning, which lets people and tools work on one part of a system at a time.

Modular reasoning

Modularity and Modular Reasoning

Languages and interfaces that let developers reason about one part of a program at a time.

Key papers: Ptolemy, a language with typed events and Panini, a concurrent programming model.

Learn more →

Software at scale

Analyzing Software at Scale, with Boa

Boa lets a software analysis be written once and run across very large collections of open-source code.

Key papers: Boa, mining ultra-large code corpora and Collective Program Analysis.

Learn more →

Dependable AI

Modular and Dependable AI

Decomposing trained models into modules, and using software engineering to make AI-enabled systems more reliable.

Key papers: DeepLocalize, fault localization for deep learning and On Decomposing a network into modules.

Learn more →

Explore our research in depth →

News

RefineAct Accepted at ASE 2026

June 19, 2026

Keeping AI Agents Within the User’s Intent Our paper “RefineAct: Automatic Runtime Verification of LLM Agent Actions” has been accepted at the 41st IEEE/ACM International...

Links: Read the paper

We Are Hiring a Postdoctoral Fellow

June 16, 2026

Postdoctoral Fellow Position Open The Laboratory for Software Design is hiring one or more postdoctoral fellows in the Department of Computer Science at Tulane University,...

Links: Apply on Tulane's career site

More News

Toward more dependable software and AI

What we work on

Modularity and Modular Reasoning

Analyzing Software at Scale, with Boa

Modular and Dependable AI

Recent Papers

RefineAct: Automatic Runtime Verification of LLM Agent Actions

Mock Deep Testing: Toward Separate Development of Data and Models for Deep Learning

An LLM-Based Agent-Oriented Approach for Automated Code Design Issue Localization

µPRL: A Mutation Testing Pipeline for Deep Reinforcement Learning based on Real Faults

IRepair: An Intent-Aware Approach to Repair Data-Driven Errors in Large Language Models

Data-Driven Evidence-Based Syntactic Sugar Design

Are Prompt Engineering and TODO Comments Friends or Foes? An Evaluation on GitHub Copilot

Inferring Data Preconditions from Deep Learning Models for Trustworthy Prediction in Deployment

Towards Understanding Fairness and its Composition in Ensemble Machine Learning

Decomposing a Recurrent Neural Network into Modules for Enabling Reusability and Replacement

Fairify: Fairness Verification of Neural Networks

Fix Fairness, Don’t Ruin Accuracy: Performance Aware Fairness Repair using AutoML

Design by Contract for Deep Learning APIs

Mutation-based Fault Localization of Deep Neural Networks

Manas: Mining Software Repositories to Assist AutoML

DeepDiagnosis: Automatically Diagnosing Faults and Recommending Actionable Fixes in Deep Learning Programs

News

RefineAct Accepted at ASE 2026

We Are Hiring a Postdoctoral Fellow

NSF Funds Our Work on Fault Localization for Deep Learning

What if a neural network were built from parts you could reuse and replace?

Making machine learning fair, from the data to the guarantee

What you can learn by reading all the world's open-source code at once

Why deep learning needs its own debugging tools

Dr. Sayma Sultana has joined the Laboratory of Software Design at Tulane University as a Postdoctoral Fellow.

Can you fix a large language model without breaking everything else it does? Our FSE 2025 paper fixes it precisely.