mohsen@ncsu:~$ whoami

* a3f9e21 (HEAD → main, origin/main)

Mohsen EsfandyariDoulabi

Ph.D. Student · NC State University

Author: Mohsen EsfandyariDoulabi <mesfand@ncsu.edu>

Date:   Ph.D. student, SWAT Lab · NC State University

mohsen@ncsu:~$ cat bio.txt

I’m a Ph.D. student in Computer Science at North Carolina State University, part of the Software Analysis and Testing (SWAT) Lab, advised by Prof. Marcelo d’Amorim.

I work on optimizing LLM agents, and build agentic, LLM-driven testing pipelines that generate test programs, triage crashes and minimize reproducers - applied to LLVM, GCC and WebAssembly runtimes, where they have surfaced 190+ bug reports across compilers, WebAssembly runtimes and deep learning libraries.

Before my Ph.D.: five years in industry — software engineering and systems programming, followed by security engineering.

192 bug reports across compilers, WebAssembly runtimes, and deep learning libraries
  • Clang / LLVM

    Clang front end, LLVM middle and back end

    75
  • GCC

    GNU Compiler Collection (Bugzilla)

    54
  • WebAssembly runtimes

    Wasmtime, WAMR, WasmEdge, Wasmer

    21
  • Deep learning libraries

    JAX, PyTorch

    42

mohsen@ncsu:~$ ls skills/

  • Python
  • C / C++
  • Go
  • LLVM
  • GCC
  • WebAssembly
  • Fuzzing
  • Differential Testing
  • Static & Dynamic Analysis
  • LLMs & Agentic AI
  • SWE-bench Verified
  • JAX / PyTorch
Mohsen EsfandyariDoulabi
Raleigh, NC

mohsen@ncsu:~$ cat research.md

* 7c14b8d

Research

--- a/research.md
+++ b/research.md

Predicting the Run, Not the Task: Failure Prediction for LLM Agents

Introduces a run-level failure-prediction framework for LLM agents, together with a budget-allocation policy that attains 3.1× the throughput of the prevailing approach. Prior evaluations rest on a metric dominated by between-task variation: it ranks tasks by relative difficulty rather than estimating whether a given run will fail, and its discriminative power degrades further as model capability increases.

Do Community-Authored Agent Skills Improve Automated Program Repair?

Presents a controlled evaluation of whether community-authored agent skills improve an LLM coding agent’s ability to repair real defects. A containerized harness over SWE-bench Verified isolates skill exposure as the sole treatment, against a matched baseline and with per-run replication for variance estimation — converting a widely held assumption in the agent-tooling community into a measurable result.

Agentic Program Analysis

Employs large language models as reasoning components within program-analysis pipelines: proposing candidate programs, triaging failures and reducing crashes to minimal reproducers. Implemented over agentic frameworks such as LangChain and applied to Clang/LLVM and GCC, substituting agent-driven synthesis for hand-written test cases.

Compiler Fuzzing

Coverage-guided and differential testing that synthesizes adversarial programs to expose crashes and miscompilations in Clang/LLVM and GCC, including an examination of what coverage guidance actually contributes to a compiler fuzzer and where its standard assumptions fail.

WebAssembly Runtimes

Large-scale differential testing across Wasmtime, WAMR, WasmEdge and Wasmer: crash discovery, seed categorization and cross-implementation comparison. Has produced 190+ bug reports across compilers, WebAssembly runtimes and deep learning libraries, with fixes contributed upstream, alongside work on reusing existing regression suites to reach code paths new tests do not.

Automated Software Testing

Reproducible, measurable test-generation pipelines — the infrastructure that converts a one-off bug hunt into a standing experiment. Spans metamorphic testing for game physics and controlled evaluation of coding agents on SWE-bench Verified.

mohsen@ncsu:~$ cat news.md

* e02da55

News

--- a/news.md
+++ b/news.md

  • Left Behind, Not Forgotten: Reusing Regression Tests to Find Bugs in WebAssembly Runtimes was accepted with minor revision at IEEE TSE.

  • Three papers under review — two at ASE 2026 and one at IEEE TSE.

  • Adaptive Ensemble Learning for Software Defect Prediction appeared at ICCKE 2025.

  • Joined the SWAT Lab as a Research Assistant, working on compiler fuzzing and WebAssembly runtime analysis.

  • Started my Ph.D. in Computer Science at NC State University.

mohsen@ncsu:~$ cat publications.md

* 1f6c9b3

Publications

--- a/publications.md
+++ b/publications.md

  1. Left Behind, Not Forgotten: Reusing Regression Tests to Find Bugs in WebAssembly Runtimes

    I. Hayat, M. EsfandyariDoulabi, G. Merabishvili, M. d'Amorim

    IEEE Transactions on Software Engineering (TSE) · 2027

  2. Demystifying Coverage-Guided Compiler Fuzzing

    M. EsfandyariDoulabi, I. Hayat, L. Liu, M. d'Amorim

    IEEE/ACM International Conference on Software Engineering (ICSE) · 2027

    Under review
  3. MetaGlitch: Physics-Grounded Metamorphic Testing for Video Game Testing

    G. Merabishvili, M. EsfandyariDoulabi, M. d'Amorim

    IEEE/ACM International Conference on Software Engineering (ICSE) · 2027

    Under review
  4. Adaptive Ensemble Learning for Software Defect Prediction: A Dynamic Weighted Hybrid Model Using SVM, DT, and ANFIS-PSO

    M. EsfandyariDoulabi, A. EsfandyariDoulabi, J. Khaligh

    International Conference on Computer and Knowledge Engineering (ICCKE) · 2025

mohsen@ncsu:~$ cat experience.md

* b84e70a

Experience

--- a/experience.md
+++ b/experience.md

Feb 2024 – Jul 2024

System Developer

Behin Afzar · Tehran, Iran

Engineered a firewall in C for IPv4, TCP/UDP, DNS and HTTP traffic, with a domain/CIDR rule engine that compiles matches into iptables policies and emits structured audit trails for compliance review.

cnetworkingdnshttpfirewalliptables

Jun 2021 – Feb 2024

Security Engineer

ShahrDI · Tehran, Iran

Automated vulnerability management end to end in Python and Bash including CVE tracking and patch-compliance verification, with anomaly detection and audit-grade logging relied on in regulatory review; led production hardening and ran secure-coding training for development teams.

securityautomationpythonhardeningcve

May 2020 – Jun 2021

SIEM Engineer

APK Group · Tehran, Iran

Detection engineering for a security operations center: authored SIEM use cases, IOC extraction parsers and analyst dashboards in Python, plus correlation logic across heterogeneous log sources that suppressed false positives and reconstructed complete attack chains.

siemdetectioniocmalwarepython

Apr 2019 – May 2020

Network Engineer

Shatel · Tehran, Iran

Managed Linux and Windows server estates and core network services (DNS, DHCP, VPN) for ISP subscribers, with scripted diagnostics and repeatable incident playbooks that cut escalation volume.

networkinglinuxdnsdhcpvpn

mohsen@ncsu:~$ cat academic.md

* 5d3af19

Academic

--- a/academic.md
+++ b/academic.md

2025–present

Research Assistant — SWAT Lab

North Carolina State University · Prof. Marcelo d'Amorim

Works on the reliability and efficiency of LLM agents and the systems they run on. Designs evaluation frameworks and benchmarking infrastructure that predict when agent runs fail and allocate compute accordingly, with controlled experiments over SWE-bench Verified in containerized harnesses. Builds automated testing pipelines — coverage-guided fuzzing, differential testing, and agentic, LLM-driven test generation — across compilers (Clang/LLVM, GCC), WebAssembly runtimes, and deep learning libraries, producing 190+ bug reports with fixes contributed upstream. Works in Python, C/C++, and Docker across large open-source codebases.

Fall 2026

CSC 474 — Network Security

North Carolina State University

Held office hours and graded assignments and projects covering network attacks, defences and applied cryptography.

Spring 2026

CSC 326 — Software Engineering

North Carolina State University

Supported team software-engineering projects: ran office hours, reviewed designs and graded iterative deliverables.

Spring 2025

CSC 574 — Computer & Network Security

North Carolina State University

Held office hours, graded assignments and projects, helped students debug their code, and clarified rubric expectations.

Fall 2024

CSC 216 — Software Development Fundamentals

North Carolina State University

Ran labs and office hours, graded programming assignments, and supported students with object-oriented design, unit testing, Git workflows, and debugging.

2019–2022

Operating Systems · Artificial Intelligence · Internet Engineering

Azad University

Taught and supported three undergraduate courses over three years, running sessions and grading for Operating Systems, Artificial Intelligence, and Internet Engineering. Voted best teaching assistant by students in 2022.

mohsen@ncsu:~$ cat skills.md

* c917e4b

Skills

--- a/skills.md
+++ b/skills.md

Languages

CC++PythonGoAssemblyBash

Testing & Analysis

FuzzingDifferential TestingStatic & Dynamic AnalysisWebAssembly Runtime AnalysisSWE-bench Verified

AI & Agents

LangChainHugging FaceAgentic FrameworksLLM-powered Test GenerationJAXPyTorch

Systems, Cloud & Web

Linux (RHEL/Debian)Windows ServerDockerAWSVMwareQEMUGitJavaScriptPHPRESTGraphQL

mohsen@ncsu:~$ cat education.md

* 2ab6f08

Education

--- a/education.md
+++ b/education.md

Aug 2024 – present

Ph.D. in Computer Science

North Carolina State University · Raleigh, NC

Software Analysis and Testing (SWAT) Lab

Advisor: Prof. Marcelo d'Amorim

GPA 3.7/4.0

Aug 2024 – Sep 2026

M.S. in Computer Science (en route)

North Carolina State University · Raleigh, NC

GPA 3.7/4.0

Nov 2017 – Apr 2022

B.Sc. in Computer Engineering

Azad University · Tehran, Iran

GPA 3.7/4.0

mohsen@ncsu:~$ cat awards.md

* 9e5c31d

Awards

--- a/awards.md
+++ b/awards.md

  • 2022

    Best Teaching Assistant

    Awarded by student vote.

    Azad University

  • 2022

    Outstanding Computer Engineering Student

    Top student across the final two semesters by semester GPA.

    Azad University

  • 2021

    Top 25 Computer Engineering Students

    Ranked on academic work and overall GPA.

    Azad University

  • 2017

    Top 7% Nationwide

    University entrance exam, Mathematics and Physics track.

    Iran National University Entrance Exam

  • 2016

    3rd Place — Tehran Province

    High school volleyball league.

    Tehran Province

  • 2013

    1st Place — Tehran Province

    U16 professional soccer league.

    Tehran Province