Sajed Jalil, Suzzana Rafi, Thomas D. LaToza, Kevin Moran, Wing Lam, ChatGPT and Software Testing Education: Promises & Perils, 2023, IEEE International Conference on Software Testing Verification and Validation Workshop, ICSTW) 10.1109/ICSTW58534.2023.00078    pdf    cite    dataset & code

Sajed Jalil, The Transformative Influence of Large Language Models on Software Development, 2023, (arXiv Pre-print)  https://doi.org/10.48550/arXiv.2311.16429    pdf    cite 

Richa Gupta, Maisha Farzana, Sajed Jalil, Thomas LaToza, A comparison of LLM and human answers to StackOverflow programming questions 2023, Journal of Student-Scientists' Research, https://doi.org/10.13021/jssr2023.3896   cite 

ChatGPT Performance Analysis | Python, Pandas, OpenAI, Playwright, Open coding
Conducted qualitative research to evaluate the efficacy of ChatGPT in addressing software development queries on Stack Overflow. Assessed the accuracy and consistency of ChatGPT’s responses to standard software testing textbook questions. Analyzed ChatGPT’s strengths and weaknesses as an AI pair programming tool.

Data Science Pipeline Detector | Python, Pandas, Jupyter Plugin, SQL
Developed a tool to detect and report various stages in data science workflows, such as data wrangling and model training.
Utilized Abstract Syntax Tree (AST) for parsing and pinpointing key terms and pipeline configurations.

ML Feed for Jupyter Lab | Python, Jupyter Plugin, SQL
Designed a Jupyter Lab plugin that recommends data science coding best practices through static/dynamic analysis, deep learning, and user interaction.
Derived best practices by mining top-rated notebooks (50+ votes) on kaggle.com.

Software defect prediction by DL and ML algorithms: A comparative study | Python, Scikit, Java
Many techniques have been developed till today to predict defects. These approaches include both classical machine and deep learning techniques. Deep learning provides better accuracy than classical machine learning over a larger dataset. In this paper, we make a comparative study to measure the accuracy and execution time of both machine learning and deep learning algorithms to predict software defects. The experimental result shows that deep learning performs significantly better in terms of accuracy.

C Path Finder ( Undergraduate Completion Project ) | Java, JavaFX, C, SQL, Z3
A Java-based automated unit test case generation tool. This uses self-implemented symbolic execution methods to generate test cases for all possible paths in C source code and also to find unreachable paths (dead code). The symbolic execution technique is exhaustive in nature and bounded by time constraints. This tool also implements memorized execution techniques to skip previously computed test cases. The tool is developed from the inspiration of the tool KLEE.

Insight on Women Empowerment using BIHS Harmonized Dataset  | Python, Scripting
A combination of supervised and unsupervised approaches to get an insight into the relationship between women empowerment and food security in Bangladesh. It was found that women’s empowerment plays a vital role in ensuring family food security.

A Study on the Origin & Evolution of God Classes on Java Systems | Java, Symbol Solver
This study was conducted on 10 open-source Java-based projects from GitHub. For each project, more than 5 versions of source code were analyzed with ensemble learning techniques to find insight into the origin and evolution of God class and potential future God classes. A new anti-pattern named Demi-God class was proposed and analysis was done on how they can be potential God classes in the future.