Publications

Research contributions

Peer-reviewed publications and preprints — with open-source code for every paper.

AcceptedIEEE AICCSA 20252025

Z-Pruner: Post-Training Pruning of Large Language Models for Efficiency without Retraining

Md. Samiul Basir Bhuiyan, Md. Sazzad Hossain Adib, Mohammed Aman Bhuiyan, Muhammad Rafsan Kabir, Moshiur Farazi, Shafin Rahman, Nabeel Mohammed

This paper introduces Z-Pruner, a novel post-training pruning technique for Large Language Models that achieves significant efficiency improvements without requiring retraining. Our method addresses the computational challenges of deploying LLMs by strategically removing redundant parameters while maintaining model performance. Extensive experiments demonstrate that Z-Pruner can reduce model size and inference time substantially while preserving the quality of generated outputs.

Under ReviewarXiv preprint2026

Bangla-WhisperDiar: Fine-Tuning Whisper and PyAnnote for Bangla Long-Form Speech Recognition and Speaker Diarization

Mohammed Aman Bhuiyan, Md. Sazzad Hossain Adib, Samiul Basir Bhuiyan, Amit Chakraborty, Aritra Islam Saswato, Ahmed Faizul Haque Dhrubo, Mohammad Ashrafuzzaman Khan

Fine-tunes Whisper for long-form Bangla ASR and PyAnnote's segmentation model for speaker diarization, tackling diverse acoustic conditions and speaker variability in real-world recordings. Achieves a 0.2441 Word Error Rate and a 0.2392 Diarization Error Rate on the test set, notably improving over pretrained baselines.

Under ReviewarXiv preprint2026

Reg2Bangla: An End-to-End Regional Speech Standardization

Samiul Basir Bhuiyan, Md. Sazzad Hossain Adib, Mohammed Aman Bhuiyan, Amit Chakraborty, Aritra Islam Saswato, Ahmed Faizul Haque Dhrubo, Mohammad Ashrafuzzaman Khan, Mohammad Abdul Qayum

Fine-tunes a Whisper-based ASR model on 3,350 dialectal recordings to transcribe twenty regional Bangladeshi dialects into standard Bangla text, with KenLM post-processing and KV-cache optimization for faster inference. Delivers strong performance despite substantial pronunciation and vocabulary variation across dialects.

Interests

Research interests

Current focus areas across applied AI and machine learning.

Applied AI and Machine Learning

Focused on developing efficient and practical AI solutions for real-world applications

Temporal Action Localization in Videos
Efficient LLM Deployment (Pruning, Quantization)
Retrieval-Augmented Generation (RAG) Systems
Small Language Models for Edge AI
AI-Integrated Web Interfaces
Cloud-Ready ML Pipelines

Focus

Research focus areas

The domains where my academic work concentrates.

Efficient AI Systems

Model compression and optimization — pruning and quantization methods that preserve performance while cutting computational cost.

  • Post-training pruning of Large Language Models
  • Model efficiency without retraining
  • Deployment optimization for edge devices

Computer Vision

Temporal action localization and video understanding through transformer-based architectures and novel training methodologies.

  • Real-time action classification in videos
  • Temporal segmentation with transformers
  • State-of-the-art results on benchmark datasets

Get in touch

Interested in the research?

I'm actively exploring efficient AI deployment, multimodal learning, and practical LLM applications — and always open to research discussions.