Md. Shohanur Islam Sobuj

About Me

I am a Machine Learning Engineer at Anymate Me. My research lies in Biomedical Image Processing, Machine Learning, and Natural Language Processing. I am a Research Associate with journals and conferences such as ICLR, EMNLP, MDPI, and IEEE. I am always open to collaboration. Feel free to drop me an email. :-)

📫 How to reach me: LinkedIn - ShohanurSobuj or
📝

Education

Hajee Mohammad Danesh Science and Technology University, Dinajpur, Bangladesh
Electrical and Electronic Engineering
Thesis Title: Diagnosis of Diabetic Retinopathy Based on Convolutional Neural Networks Using Optical Coherence Tomographic Images

Work Experience

Machine Learning Engineer
Business Automation Ltd., Dhaka, Bangladesh
November 2023 - October 2024

⚡ Developed and deployed MLOps-driven ML solutions to streamline business processes and enhance decision-making.
⚡ Built scalable ML infrastructure and microservices for efficient model training, deployment, and monitoring in production.
⚡ Managed CDC with MySQL using Debezium, Kafka, and Zookeeper for real-time data streaming.
⚡ Delivered SmartRemarks for automated comment suggestions and feedback organization via historical data analysis.
⚡ Built APIs for seamless integration, including an OCR-based TIN certificate validation system.
⚡ Led intelligent automation tools like SDG data chatbots and AI-driven decision support systems.
⚡ Implemented secure digital signature solutions to ensure document authenticity.
⚡ Automated process status selection with sentiment analysis of desk remarks.
Machine Learning Engineer
Anchorblock Technology LLC, Dhaka, Bangladesh
May 2022 - October 2023

⚡ Developed and integrated large-scale distributed machine learning systems throughout their lifecycle, utilizing cutting-edge technology.
⚡ Proficient in implementing MLOps principles, including CI/CD for ML, infrastructure management, and scalability optimization.
⚡ Hands-on experience in developing and deploying machine learning models using best practices for production environments.
⚡ Designed and created APIs for ML/AI models, facilitating easy integration and usage by other applications.
⚡ Built diverse bots, including Monty (Discord and Telegram) for spam filtering and user monitoring, Arthur (knowledgeable bot), Finix (banking chatbot), and Magic Bookkeeping (AI-powered transaction categorization system).

Research Interest

Natural Language Processing (NLP)
Machine Learning
Computer Vision
Bioinformatics

Artificial Intelligence
Data Science
Deep Learning
Data Mining

News

[Feb. 2024] "L-TUNING: Synchronized Label Tuning for Prompt and Prefix in LLMs" accepted at Tiny Papers @ ICLR 2024.
[Feb. 2024] "Leveraging Pre-trained CNNs for Efficient Feature Extraction in Rice Leaf Disease Classification" accepted @ iCACCESS 2024.
[Oct. 2023] "Contrastive Learning for Universal Zero-Shot NLI with Cross-Lingual Sentence Embeddings" accepted at EMNLP-2023.
[Oct. 2022] Thrilled to receive the prestigious Reviewer Certificate from Springer Nature, recognizing contributions in upholding scientific research quality and integrity.
[Mar. 2022] Our paper about Word Embedding Model for Transfer Learning accepted to MDPI applied science 2022.
[Aug. 2021] Our paper about Handcrafted Feature Extraction Techniques for Bangla Handwritten Digit Recognition is accepted to ICECIT 2021.
[Jun. 2021] Our paper about Bangla Language Model is accepted to ICIRCA 2021.

Publications [Google Scholar]

Published

Leveraging Pre-trained CNNs for Efficient Feature Extraction in Rice Leaf Disease Classification
Md. Shohanur Islam Sobuj, Md. Imran Hossen, Md. Foysal Mahmud, and Mahbub Ul Islam Khan
2024 International Conference on Advances in Computing, Communication, Electrical, and Smart Systems (iCACCESS)
[Paper Link] [PDF] [Code] [BibTex]
L-TUNING: Synchronized Label Tuning for Prompt and Prefix in LLMs
Md Kowsher, Md. Shohanur Islam Sobuj, Asif Mahmud, Nusrat Jahan Prottasha, and Prakash Bhat
Tiny Papers @ ICLR 2024
[Paper Link] [PDF] [Code] [BibTex]
Contrastive Learning for Universal Zero-Shot NLI with Cross-Lingual Sentence Embeddings
Md Kowsher, Md. Shohanur Islam Sobuj, Nusrat Jahan Prottasha, Mohammad Shamsul Arefin, and Yasuhiko Morimoto
EMNLP, Multilingual Representation Learning (MRL) Workshop, December 2023.
[Paper Link] [PDF] [Code] [BibTex]
An Enhanced Neural Word Embedding Model for Transfer Learning
Md. Kowsher, Md. Shohanur Islam Sobuj, Md. Fahim Shahriar, Nusrat Jahan Prottasha, Mohammad Shamsul Arefin, Pranab Kumar Dhar, Takeshi Koshiba.
Applied Sciences journal by MDPI 2022.
[Paper Link][PDF] [Code] [BibTex]
A Classical Approach to Handcrafted Feature Extraction Techniques for Bangla Handwritten Digit Recognition
Md. Ferdous Wahid, Md. Fahim Shahriar, Md Shohanur Islam Sobuj
International Conference on Electronics, Communications and Information Technology. ICECIT 2021.
[Paper Link][PDF] [Code] [BibTex] Oral Presentation
BanglaLM: Data Mining based Bangla Corpus for Language Model Research
Md. Kowsher, Md. Jashim Uddin, Anik Tahabilder, Md Ruhul Amin, Md. Fahim Shahriar, Md Shohanur Islam Sobuj
IEEE Conference on Inventive Research in Computing Applications. ICIRCA 2021.
[Paper Link][PDF] [Code] [BibTex] Oral Presentation

In Review

BERT to CNN-BiLSTM: A Hybrid Contextual Transfer Learning for Large Scale Imbalanced and Multi-label Document Classification
N.J. Prottasha, M. Kowsher, M.S.I. Sobuj, M.S. Arefin, P.K. Dhar, and T. Koshiba.
Knowledge-Based Systems, Elsevier, 2022.
An Integrated Deep Framework of CNN-BiLSTM with Attention Mechanism for Diagnosing Heart Disease
N.J. Prottasha, M.S.I Sobuj , M. Kowsher, and M.A Moni
Journal of King Saud University - Computer and Information Sciences, 2022

Preparing

Diagnosis of Diabetic Retinopathy Based on Convolutional Neural Networks Using Optical Coherence Tomographic Images.