Awards

  • 2026–2027

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

    A Course-Specific Agentic AI Learning Assistant to Support Student Learning

    I’m grateful to receive the Major CUPE 3903 Teaching Development Grant at York University to develop a course-specific Agentic AI Learning Assistant for ITEC 3020: Web Technologies and ITEC 3230: Designing User Interfaces. The system leverages modern Large Language Models (LLMs) such as ChatGPT and Gemini Pro to create a structured learning environment that helps students understand problems, review concepts, attempt their own solutions, and receive constructive feedback rather than relying on direct AI-generated answers.

    • Design a four-agent framework: Problem Interpreter, Concept Navigator, Guided Tutor, and Evaluator Agent.
    • Integrate lecture slides, lab questions, and course materials into an instructor knowledge base.
    • Use LLM-powered reasoning to provide guided hints, debugging support, and concept explanations.

    This project explores responsible integration of AI tools like ChatGPT and Gemini Pro in education while strengthening students’ conceptual understanding, debugging ability, and problem-solving skills in modern web and UI development courses.

  • 2025–2026

    Major CUPE 3903 Research Grant ($21,343)

    Optimizing Task-Specific Language Models for Multi-Domain Code Generation and Reasoning

    I’m grateful to receive the Major CUPE 3903 Research Grant at York University to support the next phase of RefactorCoderQA benchmark and multi-agent framework for reasoning-driven code generation. The project combines lightweight, edge-deployed GuideLLM with cloud-based Solver and Judge agents, enabling structured prompting, automated evaluation, and domain adaptation across Software Engineering, Data Science, ML, and NLP.

    • Build domain-adaptive MoE variants tailored to real coding tasks and constraints.
    • Scale RefactorCoderQA with new datasets RefactorCoderQA Pro, reasoning modules, and richer evaluation pipelines.
    • Advance agentic collaboration between coding and reasoning LLMs for faster, interpretable solutions.

    This funding strengthens our goal of making AI-assisted code reasoning more transparent, reliable, and useful for education and real-world developer workflows.

  • 2024–2025

    CUPE 3903 Conference Travel Fund ($5,500)

    Support for Attending the ACL, ICOA & FLLM 2025 Conferences — Vienna, Austria

    I was awarded the CUPE 3903 Conference Travel Fund to present my research at the ACL 2025, ICOA2025 and FLLM 2025 conferences in Vienna, Austria. The fund provides financial support for contract faculty to participate in major scholarly venues, covering travel, accommodation, and registration expenses.

    This recognition highlights York University’s continued support for academic excellence and global research engagement, allowing me to share and expand the impact of my work on Large Language Models, Code Generation, and Multimodal Reasoning.

  • March 2020

    DIU Research Award ($800)

    Recognition for B.Sc. thesis publication

    Awarded by Daffodil International University (DIU) for my undergraduate thesis published with Springer: “Assessing the Effectiveness of Topic Modeling Algorithms in Discovering Generic Labels with Description”, featured in Advances in Information and Communication (FICC 2020, Vol. 2). View chapter.

News

  • Accepted
    2025
    Paper Accepted to EACL Finding 2026.
    Benchmarking agentic reasoning for dashboards and interactive UIs.
  • Media
    2025
    Media coverage: RefactorCoderQA featured as innovative research at Quantum Zeitgeist.
    Highlights cloud–edge MoE design and multi-domain performance gains.
  • Accepted
    2025
    Paper submitted to IEEE Transactions on Services Computing.
    RefactorCoderQA: reasoning-driven code generation and evaluation across cloud & edge.
  • Published
    2025
    Paper accepted in Findings of ACL 2025.
    ChartQAPro: Continuing our work on multimodal reasoning and evaluation.
  • Accepted
    2025
    Two papers accepted to FLLM 2025.
    Exploring energy-efficient code generation and large-foundation-model workflows.