I am a first-year PhD student in Computer Science and Engineering at Kennesaw State University. My research focuses on multimodal human sensing: building systems that recognize human activity from heterogeneous sensors, such as mmWave radar, WiFi, RFID, vision, and wearables, and that stay reliable when some of those sensors fail, drop out, or are unavailable at deployment. More broadly, I am interested in robust, transferable, and interpretable multimodal representations grounded in physical and symbolic structure.

Before starting my PhD, I spent over five years in industry and research labs in Dhaka, Bangladesh. As a Senior Machine Learning Engineer at Silicon Orchard Ltd, I worked on using knowledge graphs to monitor and constrain failure modes of generative AI, physics-guided multimodal clustering for aircraft imagery, anomaly detection in satellite imagery that generalizes to unseen conditions, diffusion-based augmentation for low-data settings, and LLM-based agentic systems for tool learning. Earlier, at Apurba Technologies, I worked on OCR and document understanding, including learning-based layout analysis, robust text recognition across diverse document styles, and scalable inference for multimodal models, and I mentored six research assistants in an industry–academia collaboration. I began my research career at the Apurba–DIU R&D Lab, where I led the construction of a large-scale OCR dataset for a low-resource language.

My work has been published at venues including EMNLP (Industry Track), WACV, IEEE BigData, and the AAAI Symposium Series. I hold a Bachelor’s degree in Computer Science and Engineering from Daffodil International University, where I graduated magna cum laude.

• Our work “Continuous Monitoring of Large-Scale Generative AI via Deterministic Knowledge Graph Structures” has been accepted on AAAI 2025 Fall Symposium Series.
• Our team has been selected as a semi-finalist in the CivX Data to Decision Challenge: Harnessing Digital Twin Data for Intelligent Decisions.
• Our work on Advanced Tool Learning and Selection System (ATLASS) is accepted as a full paper for presentation in CISOSE 2025.
• Starting new position as a Senior Software Engineer (Machine Learning) at Silicon Orchard Ltd.
• Our work on the Enhancement of Bengali OCR is Accepted at WACV 2024, Workshop on Vision-Based Understanding for Low-Resource Languages.
• I am serving as a Program Committee for ICMLA 2024.
• Our work on OCR for Versatile Documents is Accepted at IEEE BigData 2023.
• Our work on the biggest Bangla Character and Word Level Dataset is Accepted at the EMNLP 2023 Industry Track.