EMNLP'26Selected
GraphBit: A Graph-Based Agentic Framework for Non-Linear Agent Orchestration
Authors: Y. Sarker, M. R. Ullah, M. Molla, S. Joty
Proc. Conference on Empirical Methods in Natural Language Processing (EMNLP), Industry Track,
Yeahia Sarker
Graduate Assistant @ MTSU
I am a Graduate Assistant at Middle Tennessee State University, where I am pursuing an M.S. in Engineering Technology and working on AI for cyber-physical systems: graph neural networks and attention for fault diagnosis in power grids and inverters, and LLM agents for autonomous drones.
My recent work includes GraphBit, a graph-based framework for multi-agent orchestration (EMNLP 2026), and a unified energy-based generative framework for hyperspectral image classification (IEEE ICIP 2025). Before MTSU I spent six years in industry building and leading machine learning teams. I hold a B.Sc. in Mechatronics Engineering from RUET, Bangladesh.
📢 If you are interested in collaboration, please feel free to reach out via email.
My recent work includes GraphBit, a graph-based framework for multi-agent orchestration (EMNLP 2026), and a unified energy-based generative framework for hyperspectral image classification (IEEE ICIP 2025). Before MTSU I spent six years in industry building and leading machine learning teams. I hold a B.Sc. in Mechatronics Engineering from RUET, Bangladesh.
📢 If you are interested in collaboration, please feel free to reach out via email.
Murfreesboro, TN
Research Interests
- Pretraining SLM
- Multi-Agent Orchestration
- Graph Neural Networks
- Cyber-Physical Systems
- Computational 3D Imaging
- Human-Robot Interaction
News
GraphBit paper accepted to the EMNLP 2026 Industry Track.
Preprint on cognitive graph intelligence for DDoS attack detection in next-generation networks released on arXiv.
Crossed 1,000 citations on Google Scholar.
GraphBit, a graph-based agentic framework for non-linear agent orchestration, released on arXiv.
Invited to serve as a reviewer for EMNLP 2026.
Joined Middle Tennessee State University as a Graduate Assistant in the Department of Engineering Technology.
Selected Publications
ICIP'25Selected
Energy-Based Generative Models with Morphological Attention Networks for Hyperspectral Image Classification: A Unified Framework
Authors: M. A. Hossain, Y. Sarker, W. Liu, N. Ansari
Proc. IEEE International Conference on Image Processing (ICIP), pp. 2282–2287,