About
I am a PhD Student in the Thomas J. Watson College of Engineering and Applied Science at Binghamton University ( SUNY ), which is located in the New York at Binghamton. I am working as a Research Assistant ( RA ) at Real-Time Embedded Systems Laboratory ( RTlab ) which is supervised by the professor Kyoung-Don Kang.
My research interests include enhancing security within machine learning frameworks and advancing unsupervised learning techniques. I have been recognized through scholarly publications and have led impactful projects like 'Third Eye' for assistive technology and smart systems for autonomous vehicles. Proficient in various programming languages and ML libraries, my expertise was honored with a podium placement in the Ebay ML Challenge and a scholarship for technical merit.
My academic roots trace back to Bangladesh University of Engineering and Technology, where I honed my foundation in engineering, leading to advanced research in computing.
After graduating, I embarked on a journey with Cobalt Speech and Language, a USA based company , as a Research Scientist, primarily focusing on Natural Language Processing (NLP) and Understanding (NLU) tasks. I played a key role in one of the ambitious projects of my company called “Call Summarization” , training large language models like BART, BERT, and LED , thereby gaining extensive experience with transformer models.
Most recently, I spent the summer of 2026 at eBay as an Applied Research Intern, working on region-level multimodal retrieval. On the research side, I studied how much localized retrieval ability frozen vision-language backbones already carry without any gradient updates, which led to TIP-REGION and a leakage-controlled benchmark for small-referent retrieval. In parallel, I fine-tuned a BLIP model with visual-grounding late interaction on eBay Fashion data so that fine-grained attributes survive at scoring time instead of collapsing into a single whole-image embedding. Working at industry scale taught me a great deal about large-scale experimentation, efficient use of compute, reproducibility, and the engineering that carries a research prototype toward a production system.
When not immersed in algorithms, you'll find me hiking trails, playing soccer, or savoring the thrill of a good game - activities that sharpen my focus and fuel my passion for innovation.
Recent News
August 2026
Wrapped up my summer as an Applied Research Intern at eBay. We introduced TIP-REGION, a training-free paired-exemplar method that consistently improves region-level retrieval across three heterogeneous frozen backbones (CLIP, SigLIP2, and Qwen3-VL-Embedding), together with LOCFG, a leakage-controlled benchmark for small-referent retrieval; the paper is being prepared for submission to ICLR. I also fine-tuned a BLIP model with visual-grounding late interaction on eBay Fashion data, substantially improving over the existing retrieval model.
My team also won eBay's Vibe Coding Hackathon across all intern teams. The project was recognized by eBay CEO Jamie Iannone and drew follow-up interest from the marketing team.
May 2026
Joined eBay in California as an Applied Research Intern, working on multimodal retrieval for e-commerce search.
June 2025
Our paper accepted in Sensors titled "TDA-L: Reducing Latency and Memory Consumption of Test-Time Adaptation for Real-Time Intelligent Sensing".
November 2024
Our team secured second runner-up in the eBay ML 2024 competition, where we developed a solution for extracting vehicle fitment from advertised data using LLMs such as Mistral Nemo 12B and Qwen 2.5 14B.
August 2024
Started my PhD in Binghamton University (SUNY) at school of computing under the supervision of Prof. Kyoung-Don Kang.
May 2024
Defended my Master's thesis titled "A DNN for Detecting Spotted Lanternflies Using Energy Efficient WAN" at Binghamton University (SUNY).
December 2023
Our team secured second runner-up in the eBay ML 2023 competition, where we developed a semi-supervised teacher-student framework to enhance product tag tiling performance using RoBERTa.
August 2023
Our paper accepted in ICCV 2023 titled "SSDA: Secure Source-Free Domain Adaptation".
May 2023
Joined Binghamton University (SUNY) as a Research Assistant.
March 2022
Joined Cobalt Speech and Language as a Research Scientist.
July 2021
Our paper accepted in IJCNN 2021 titled "Dynamic perturbation of weights for improved data reconstruction in unsupervised learning".
April 2021
My paper accepted in I2CT 2021 titled "A hybrid clustering pipeline for mining baseline local patterns in 3d point cloud".