Aryaz EghbaliProjectsExperience
Aryaz Eghbali

My name is Aryaz Eghbali.

I am a PhD student in the Software Lab at the University of Stuttgart, Germany.

My area of interest is automating software analysis and development. Recently, I have been working on automatically generating tests, a novel metric to evaluate similarity in source code called CrystalBLEU, and the first general-purpose dynamic analysis framework for Python called DynaPyt.

I previously worked on distributed algorithms and distributed systems.

Projects
TestPilot

Max Schäfer, Sarah Nadi, Aryaz Eghbali, Michael Pradel

An Empirical Evaluation of Using Large Language Models for Automated Unit Test Generation.

IEEE Transactions on Software Engineering (TSE 2024).

Aryaz Eghbali, Michael Pradel

DynaPyt is the first general-purpose dynamic analysis framework for Python. It provides analysis hooks of various abstraction levels, ranging from individual operations to all runtime events. Analyses in DynaPyt are also able to modify the runtime values, which allows the possibility of implementing analyses like concolic testing.

The ACM Joint European Software Engineering Conference and Symposium on the Foundations of Software Engineering (ESEC/FSE 2022).

ACM SIGSOFT Distinguished Paper Award

Aryaz Eghbali, Michael Pradel

CrystalBLEU is a novel metric for evaluating similarity between source code. It is based on BLEU (Papineni et. al 2002), with higher distinguishability.

The 37th IEEE/ACM International Conference on Automated Software Engineering (ASE 2022).

Aryaz Eghbali, Michael Pradel

An empirical study on string-related bugs in JavaScript, analyzing how they manifest, their root causes, and how they can be repaired.

The 35th IEEE/ACM International Conference on Automated Software Engineering (ASE 2020).

Aryaz Eghbali, Roger Wattenhofer

A heuristic-based modeling of the BitCoin mining hardware distribution over time.

The 3rd International Workshop on Cryptocurrencies and Blockchain Technology (CBT 2019).

Aryaz Eghbali, Philipp Woelfel

Leader election is the problem of electing a leader from a set of processes. We show that there is a lower bound on the number of remote memory references (RMRs) required for abortable leader election in a distributed system.

The 32nd International Symposium on Distributed Computing (DISC 2018).
Experience
PhD in Computer Science

2020 - present, University of Stuttgart

Supervisor: Michael Pradel
Working on software analysis tools, metrics for evaluating code generating models, and automating software engineering tasks.
Research Intern

Summer 2022, GitHub

Supervisor: Max Schaefer and Frank Tip
Worked on automated test generation
Researcher

2018 - 2019, ETH Zurich

Supervisor: Roger Wattenhofer
Worked on modeling the hardware usage by Bitcoin miners.
Data Science Intern

Summer 2018, Divar.ir

Worked on analyzing in-app chat data to detect spam and offensive messages.
M.Sc. in Computer Science

2014 - 2018, University of Calgary

Supervisor: Philipp Weolfel
Thesis on proving a lower bound for abortable leader election algorithm.
B.Sc. in Electrical & Computer Engineering

2010 - 2014, University of Tehran

Talks
Dynamically analyzing Python programs using DynaPyt
ASE tutorial
Iterative code completion using large language models
Dagstuhl Seminar on Programming Language Processing
Service
External reviewer for ISSTA 2024

Reviewer for Transactions on Interactive Intelligent Systems (TiiS)