Md. Asaduzzaman Shuvo

Md. Asaduzzaman Shuvo

Aspiring AI/ML Engineer | NLP & Vision-Language Models (VLMs) Researcher · United international University

I'm Md. Asaduzzaman Shuvo, an AI/ML researcher and engineer with a B.Sc. in Computer Science & Engineering from United International University (UIU). My research sits at the intersection of NLP and vision-language models, with a focus on low-resource languages and real-world multimodal understanding. I build datasets and benchmarks that reveal where and why LLMs fail, especially in Bangla text generation, linguistic register, and culturally grounded reasoning, and I make compact models more capable through LoRA/QLoRA, quantization, and reasoning distillation. I also apply explainable vision-language systems to environmental and public-health problems such as mosquito breeding-site analysis and floating-waste detection. I'm currently seeking research opportunities in NLP, multimodal learning, and trustworthy LLMs.

7
Papers
4
First-author
6
Venues
2
Years active
250+
Problems solved

News

Services

Reviewer5

Publications

Co-author network

Circle size shows the number of shared papers. Select a co-author to filter the list below.

Riasad AlviMd. Asaduzzaman ShuvoMd. Ahashan HabibMd. Faiyaz Abdullah SayeediMd. Adnanul IsamMahedi HasanMd. Tashin ParvezAzizul Haque NomanMd. Shafayet Hossain OviShahanur Rahman BappyMuhammad Ziaur RahmanMd Asiful IslamSwakkhar ShatabdaSuvro SarkarRiasat AzimAhsan FarabiMd Abdul Ahad MinhazIsrat KhandakerIbrahim Khalil ShantoMuhammad Nomani KabirMd. Rasel AliJannatun Noor

Showing 7 of 7

  1. WACV 2027
    Entropy-Gated Dynamic Rectification for Object Hallucination in Large Vision-Language Models

    Riasad Alvi, Md. Asaduzzaman Shuvo*, Md. Ahashan Habib, Md. Faiyaz Abdullah Sayeedi

    WACV · conference · 2026Accepted

    arXivCode
  2. EMNLP 2026
    When a Name Is Not a Name: A Benchmark Dataset and Distilled Reasoning for Culturally Entangled Bangla Homographs in Low-Resource LLMs

    Md. Asaduzzaman Shuvo*, Md. Adnanul Isam

    WiNLP · workshop · 2026Published

    PaperCode
  3. EMNLP 2026
    Polite on the Surface, Broken in Practice: A Curated Dataset for Fixing Generation and Register Failures in Low-Resource Bangla Text Generation
    EMNLP 2026

    TL;DR BLADE, a 4,196-example curated dataset, fixes Bangla LLMs' generation, structural, and honorific failures via LoRA fine-tuning, showing curated data can outperform raw model scale across automatic, human, and LLM-judge evaluations.

    Md. Asaduzzaman Shuvo*, Mahedi Hasan, Md. Tashin Parvez, Azizul Haque Noman, Md. Shafayet Hossain Ovi

    GroundLM · workshop · 2026Published

    BLADE, a 4,196-example curated dataset, fixes Bangla LLMs' generation, structural, and honorific failures via LoRA fine-tuning, showing curated data can outperform raw model scale across automatic, human, and LLM-judge evaluations.

    PaperCode
  4. CVPR 2026
    VisText-Mosquito: A Unified Multimodal Dataset for Visual Detection, Segmentation, and Textual Explanation on Mosquito Breeding Sites

    Md. Adnanul Isam, Md Faiyaz Abdullah Sayeedi, Md Asaduzzaman Shuvo*, Shahanur Rahman Bappy, Muhammad Ziaur Rahman, Md Asiful Islam, Swakkhar Shatabda

    AI4RWC · workshop · 2026Published

    PaperCode
  5. Tensymp
    ADDomics: Enhancing Multidimics Fusion with Autoencoders and Deep Neural Networks

    Md Ahashan Habib, Riasad Alvi, Suvro Sarkar, Md Asaduzzaman Shuvo*, Mahedi Hasan, Riasat Azim

    IEEE Region 10 Symposium (TENSYMP) · conference · 2025Published

    Paper

Under review2

  1. Arxiv
    WADE: A Reasoning-Annotated Benchmark for Multi-Instance Floating-Waste Grounding with Compact Vision-Language Models

    Md Asaduzzaman Shuvo*, Ahsan Farabi, Md Abdul Ahad Minhaz, Mahedi Hasan, Israt Khandaker, Ibrahim Khalil Shanto, Muhammad Nomani Kabir

    journal · 2026Under review

    arXiv
  2. ARR
    Who May Change the Policy? Authorization-State Failures in Tool-Using Agents

    Riasad Alvi, Md. Ahashan Habib, Md. Rasel Ali, Md. Asaduzzaman Shuvo*, Md. Faiyaz Abdullah Sayeedi, Jannatun Noor

    ACL Rolling Review · conference · 2026Under review

Timeline

Hover a period to highlight the papers and awards from those years.

BSc in Computer Science and EngineeringUnited International University
ResearcherUnited International University
AI/ML Engineer InternshipParallax Labs
2022
2023
★ 1st Runners up, Database management system UIU CSE Project Show
2024
2025
Tensymp★ Awarded a 50% merit-based scholarship★ 2nd Runners up, System Analysis and Design Laboratory UIU CSE Project Show★ 3rd Runners up, Microprocessors and Microcontrollers Laboratory UIU CSE Project Show
2026
WACV 2027EMNLP 2026EMNLP 2026CVPR 2026ArxivARR★ Received a WiNLP@EMNLP Travel Grant

Education

  • BSc in Computer Science and EngineeringUnited International University, Dhaka, Bangladesh
    2022 – 2025
  • ScienceGovt. K.C. Collage, Jhenaidah, Bangladesh

Experience

  1. AI/ML Engineer Internship

    Parallax Labs

    Oct – Dec 2026

    Key Areas of Expertise

    • Data Engineering & Exploratory Data Analysis
    • Text Chunking Strategies & Embedding Models
    • Vector Database Management (ChromaDB)
    • LLM API Integration & Prompt Engineering
    • RAG Evaluation: Precision, Recall & Hallucination Detection
    • NLP: Topic Modeling & Sentiment Analysis
    • API Development (FastAPI) & Performance Benchmarking

    Tools

    • Python
    • ChromaDB
    • FastAPI
    • ArXiv API
    • Groq
    • OpenRouter

  2. Researcher

    United International University

    Jan 2023 – May 2026

    • Authored 5 accepted papers at WACV 2027, CVPR 2026 Workshop and EMNLP 2026 Workshops (WiNLP, GroundLM).
    • Built and released 2 Bangla NLP benchmark datasets on culturally entangled homographs and register failures; benchmarked 10+
    • LLMs and improved compact models with QLoRA, 4-bit quantization, and cultural chain-of-thought distillation.
    • Co-developed an entropy-gated rectification method that reduces object hallucination in large vision-language models (WACV).
    • Built WADE (thesis), a reasoning-annotated benchmark and distillation framework for floating-waste grounding with compact

    VLMs; contributed to VisText-Mosquito, a multimodal detection, segmentation, and explanation dataset (CVPR 2026 Workshop)

Awards

  • Received a WiNLP@EMNLP Travel Grant
    2026
  • Awarded a 50% merit-based scholarship
    2025
  • 2nd Runners up, System Analysis and Design Laboratory UIU CSE Project Show
    2025
  • 3rd Runners up, Microprocessors and Microcontrollers Laboratory UIU CSE Project Show
    2025
  • 1st Runners up, Database management system UIU CSE Project Show
    2023