Sciunit
A command-line tool for creating reusable, self-contained "sciunits" that audit and repeat program executions exactly, regardless of the deployment environment — the reproducibility engine Floability builds on.
Raza Ahmad
Final-Year PhD Student · Computer & Information Sciences · DePaul University
I study record-and-replay techniques for interactive and distributed computing — building tools that make notebooks and HPC workflows portable, auditable, and reproducible. This builds on my background in data provenance, where I've worked on auditing, querying, efficiently storing, and protecting system-level provenance metadata. I also apply machine learning to a range of problems, ranging from predicting execution paths in provenance graphs to text and sentiment analysis and early disease detection.
About
Raza Ahmad is a final-year PhD student in Computer & Information Sciences at DePaul University, where his research focuses on record-and-replay techniques for interactive and distributed computing — building systems that make notebooks and HPC workflows portable, auditable, and reproducible. His recent work includes NBReplay, an audit & repeat framework for efficient re-execution of distributed notebook workflows, and Floability, a system for portable deployment of notebook-based scientific workflows across HPC clusters. His 2022 paper introducing FLINC, a method for creating reproducible notebook containers through application virtualization, received the Best Student Paper Award at the IEEE International Conference on e-Science.
Before returning to DePaul as a PhD student, Raza spent several years working across research and industry. As an International Research Fellow at SRI International, he contributed to SPADE, an open-source whole-system provenance auditing framework, developing its querying and storage architecture across multiple database backends. Earlier, as a Research Associate at Lahore University of Management Sciences, he worked on applied machine learning, natural language processing, and information retrieval. He holds a Master's degree in Computer Science from DePaul University and a Bachelor's degree in Computer Science from FAST–National University of Computer and Emerging Sciences, Lahore.
Raza has taught courses including Database Systems, Object-Oriented Analysis and Design, Assembly Language Programming and Computer Architecture, and Theory of Automata. He has also served as a reviewer for Distributed and Parallel Databases and Future Generation Computer Systems.
Research
Efficiently Reproducing Distributed Workflows in Notebook-based Systems
Backpacks for Notebooks: Enabling Containerized Notebook Workflows in Distributed Environments
Improving Reproducibility of Interactive Notebooks Using Application Virtualization
Accurate Path Prediction of Provenance Traces
Comparing Containerization-Based Approaches for Reproducible Computational Modeling of Environmental Systems
Improving Reproducibility of Geoscience Models with Sciunit
Analyzing Discrepancies in Whole-Network Provenance
Reproducible Notebook Containers Using Application Virtualization
Sciunit: A Reproducible Container for EarthCube Community
Digging into Big Provenance (with SPADE)
Practical Provenance Privacy Protection
Content-defined Merkle Trees for Efficient Container Delivery
An Approach for Open and Reproducible Hydrological Modeling Using Sciunit and HydroShare
Discrepancy Detection in Whole Network Provenance
PROV-CRT: Provenance Support for Container Runtimes
Streaming Provenance Compression
Aspect Based Sentiment Analysis for Large Documents with Applications to US Presidential Elections 2016
Accurate Detection of Automatically Spun Content via Stylometric Analysis
Decision-tree Inspired Classification Algorithm to Detect Tuberculosis (TB)
Malware Slums: Measurement and Analysis of Malware on Traffic Exchanges
Projects
Floability
A framework for portable deployment of notebook-based scientific workflows — packages code, environment, and data into a self-contained "backpack" that deploys across HPC clusters via TaskVine and captures fully reproducible runs via Sciunit.
SPADE
An open-source whole-system provenance auditing framework that collects, filters, stores, and queries provenance metadata across distributed environments — the subject of the CACM/ACM Queue piece on querying "big provenance."
NBReplay
Audit/repeat kernels that incrementally checkpoint notebook state and selectively replay only the distributed tasks affected by a change — cutting re-execution time by up to 46× in early evaluation on Dask/TaskVine workflows.
FLINC
Reproducible notebook containers built through application virtualization — automatically captures a notebook's full dependency and execution environment so it can be replayed exactly as originally run, without Docker.