Projects

Open-source tools and research prototypes focused on AI-assisted software engineering, dependency resolution, program analysis, and automated repair.

tooool: Transforming Software Engineering Problems into SAT/SMT

University of Windsor / Independent | 2024 – Ongoing

Core passion project: Converts complex SE challenges (dependency solving, configuration, Boolean reasoning) into Boolean Satisfiability (SAT) problems solvable by high-performance SMT solvers. Bridges classical constraint solving with modern software engineering pain points.

Tools & Techniques: Python, Z3/PySMT, SAT/SMT solvers, Constraint modeling, PyPI/Maven integration

Status: Active prototype – expanding to more SE domains

Link: GitHub: tooool

PyHealer: LLM- and RAG-enabled Agentic Repair of Python Execution Failures

University of Windsor | 2025 – 2026

Agentic AI system using LLMs + Retrieval-Augmented Generation (RAG) to diagnose and auto-repair Python runtime errors from missing deps, env mismatches, config issues, or code faults. Aims to reduce manual debugging in research/scientific Python workflows.

Tools & Techniques: Python, LLMs (e.g., GPT/Llama), LangChain/LlamaIndex-style RAG, Agent frameworks, Error parsing, Virtual env automation

Link: Zenodo Record (dataset/tool artifact)

PyHealer architecture or demo screenshot

SMTpip: Solver-Based Dependency Management for Python

University of Windsor (M.Sc. Thesis component) | 2024 – 2025

Prototype tool that translates Python dependency requirements into SMT constraints for optimal, conflict-free resolution — outperforming greedy pip in complex scenarios. Directly tied to thesis on better dependency ecosystems.

Tools & Techniques: Python, SMT solvers (Z3), PyPI API, Constraint satisfaction, Version modeling

Link: GitHub: SMTpip

SMTpip interface or resolution example

Custom AST Parsers for Python & Java

University of Windsor | 2023 – 2024

Built extensible parsers and visualizers for Abstract Syntax Trees to support static analysis, code transformation, and research tooling in program comprehension.

Tools: Python ast module, Eclipse JDT (Java), Graph visualization

Links:
• Python AST Parser
• Java AST Parser

VTA Bot: Virtual Teaching Assistant

BRAC University (Undergrad Thesis) | 2021 – 2022

Conversational AI bot powered by NLP and deep learning to answer student queries, explain concepts, and assist with coursework — early exploration of AI in education.

Tools & Techniques: NLTK, PyTorch, Intent classification, Dialogue management, Data collection

Link: GitHub: VTA Bot

VTA Bot conversation interface

All projects are open-source where possible. Interested in collaborating on agentic AI for SE, SMT applications, or dependency tools? Reach out via LinkedIn or GitHub!