Adjunct Faculty • ML/NLP Researcher

Shadikur Rahman

I build practical AI systems for code quality, chart understanding, and real-world ML applications. Seeking a PhD position as a domestic student and ML/NLP roles.

See Publications Featured Projects CV
Portrait of Shadikur Rahman

About

I am an Adjunct Faculty of Information Technology at York University, Canada and a Research Associate at Algoma University, where I build benchmarks and frameworks for Large Language Models (LLMs), code generation, and multimodal reasoning. Previously, I worked as a Software Engineer at Samsung R&D Institute Bangladesh Ltd. and completed my M.Sc. in Computer Science at York University under the supervision of Prof. Enamul Hoque Prince , focusing on LLM-assisted code generation based on code reviews.

My research bridges software engineering, NLP, and multimodal ML to improve the reliability and interpretability of AI systems for developers and analysts. I design benchmarks and agentic pipelines that connect language, code, and vision—advancing model reasoning for chart understanding, code generation, and human–AI collaboration.

Recent projects include RefactorCoderQA, ChartQAPro, and DashboardQA with publications in venues such as ACL Findings, TSC, EMSE, ESEM, IEEE REW, and FLLM. I am actively seeking Ph.D. opportunities and research collaborations in ML, NLP, and multimodal AI.

Education

MSc in Computer Science at York University

Jan 2022 – Dec 2023
Toronto, Canada
GPA: 3.70 / 4.0

Thesis: Enhancing code review for improved code quality with language model-driven approaches.

BSc in Software Engineering at Daffodil International University (DIU)

Jan 2015 – Dec 2018
Dhaka, Bangladesh
GPA: 3.54 / 4.0

Thesis: Polynomial Topic Distribution with Topic Modeling for Generic Labeling.

Experience

Adjunct Faculty of Information Technology at York University

May 2024 – Present • Toronto, Canada
  • ITEC 3020 – Web Technologies: HTML, CSS, Bootstrap, JavaScript, Node.js, and MySQL.
  • ITEC 3230 – Designing User Interfaces: user-centered design, prototyping, and usability evaluation.
  • Designing and delivering interactive lectures and hands-on coding sessions.
  • Guiding students in developing real-world projects and improving their problem-solving skills.

Research Associate at Algoma University

Nov 2024 – Present • Brampton, Canada
  • Developed RefactorCoderQA, a benchmark for multi-domain code reasoning with LLMs.
  • Designed a multi-agent LLM framework (GuideLLM–SolverLLM–JudgeLLM) for structured reasoning.
  • Fine-tuned RefactorCoder-MoE using QLoRA on DeepSeek-Coder-7B, achieving 76.84% accuracy.
  • Built JudgeLLM using GPT-4 for automated accuracy and clarity evaluation.
  • Performed latency, ablation, and human–AI evaluation studies across SE, DS, ML, and NLP domains.

Research Assistant at Intelligent Visualization Lab, York University

Jan 2024 – Present • Toronto, Canada
  • Contributed to ChartQAPro and DashboardQA benchmarks for multimodal and GUI-based reasoning.
  • Curated real-world charts and dashboards with expert-authored QA pairs.
  • Developed evaluation pipelines for VLMs and agentic models using GPT-4o, Gemini, and Claude.
  • Analyzed grounding, planning, and reasoning gaps in multimodal LLM performance.
  • Supported dataset design, experimentation, and benchmarking for vision–language understanding.

Graduate Research Assistant at York University

Jan 2022 – Jan 2024 • Toronto, Canada
  • Developed an application to identify and recommend similar code reviews to improve code quality.
  • Fine-tuned a BERT model for code review classification (96% F1-score).
  • Retrieved relevant Stack Overflow data using chunking and NER for contextual insights.
  • Built an LLM chatbot using Llama 2 to assist developers in issue resolution.

Research Assistant at Umm Al-Qura University

Jun 2020 – Jul 2021 • Saudi Arabia (Remote)
  • Implemented a CNN-based technique to classify COVID-19 from X-ray images, comparing VGG and ResNet models.
  • Proposed an optimized CNN model achieving 97% accuracy in distinguishing COVID-19 cases.

Software Engineer at Samsung R&D Institute Bangladesh Ltd.

May 2019 – Nov 2019 • Dhaka, Bangladesh
  • Analyzed code reviews from GitHub repositories to identify coding issues using Machine Learning and NLP.
  • Applied SVM with TF-IDF for classifying ambiguous code reviews (90% accuracy, 92% F1-score).
  • Developed a developer-assistance tool using Django and Django REST Framework in Python.

Awards

Major CUPE 3903 Teaching Development Grant ($21,343)

2026–2027
York University

Funding for the project “A Course-Specific Agentic AI Learning Assistant to Support Student Learning”, focused on developing a four-agent AI framework that guides students through problem understanding, concept review, guided solution attempts, and constructive feedback for courses ITEC 3020 and ITEC 3230.

Major CUPE 3903 Research Grant ($21,343)

2025–2026
York University

Funding for the project “Optimizing Task-Specific Language Models for Multi-Domain Code Generation and Reasoning”, supporting the next phase of the RefactorCoderQA benchmark and multi-agent LLM framework for reasoning-driven code generation.

CUPE 3903 Conference Travel Fund ($5,500)

2024–2025
York University

Awarded to present research at ACL, ICOA, and FLLM 2025 in Vienna, Austria, showcasing work on Large Language Models, Code Generation, and Multimodal Reasoning.

DIU Research Award 2020 ($800)

March 2020
Daffodil International University

Recognition for B.Sc. thesis published with Springer: “Assessing the Effectiveness of Topic Modeling Algorithms in Discovering Generic Labels with Description” (FICC 2020, Vol. 2).

Research Projects

Optimizing Task-Specific Language Models for Multi-Domain Code Generation and Reasoning (RefactorCoderQA Pro)

2025–2026
Major CUPE 3903 Research Grant, York University
  • Developing a large-scale benchmark and multi-agent framework (Guide–Solver–Reviewer–Judge) for reasoning-driven code generation.
  • Evaluating LLMs across software engineering, data science, and NLP domains to study trade-offs in accuracy, clarity, and efficiency on cloud and edge deployments.

Review2Code: Benchmarking LLM-Driven Code Generation from Code Review Comments

2025–2026
Major CUPE 3903 Research Grant, York University
  • Designing a pipeline to translate human code review feedback into optimized, executable code using instruction-tuned LLMs.
  • Benchmarking models for actionable fix generation while maintaining logical and syntactic consistency.

UI2Code-Real: Bridging Visual Web Design and Front-End Code Generation from UI Scratch

2025–Present
York University
  • Creating a realistic design-to-code benchmark using 100+ full-stack student projects (HTML, CSS, Bootstrap, Node.js).
  • Evaluating multimodal LLMs (GPT-4o, Claude 3.5, Gemini 1.5 Pro) on translating UI screenshots into accurate, semantically rich front-end code.

Featured Research

LLMs Code Generation LLMs Reasoning

RefactorCoderQA

Benchmark + multi-agent pipeline (Guide/Solver/Judge) for real coding tasks.

Paper Quantum News GitHub
Vision Visualization

ChartQAPro

Diverse chart QA benchmark with harder questions & robust eval protocols.

Paper GitHub Hugging Face
Agents GUI

DashboardQA

Agents for question answering on interactive dashboards.

Paper

Skills

Programming

Python C Java SQL (PostgreSQL, MySQL)

ML / DL

scikit-learn pandas NumPy TensorFlow PyTorch Keras OpenCV

NLP / LLMs

LLMs LangChain Ollama Retrieval NER LLM reasoning VLLM Quantization Fine-tuning Evaluation

Volunteer Experience

  • Reviewer

    Peer reviewer for NAACL, ACL, ESEM, and CSSE venues — evaluating submissions in NLP, software engineering, and ML.

  • Supervision

    DIU Research Lab: mentoring student researchers on projects in code generation, benchmarking, and applied NLP (proposal review, experiment design, and writing feedback).

  • Startup

    TorontoRides: collaborating on an early-stage mobility platform (service design, web stack architecture, and operations planning).

Contact

3052 Victor Phillip Dahdaleh Building, 4700 Keele Street Toronto, Ontario, Canada, M3J 1P3.

Email
shadikur@yorku.ca
Phone
+1 (416) 562-6670