I am a Master's student of IIT, University of Dhaka, and have completed my BSc in CSE from North South University. My Undergrad thesis was an NLP and Deep Learning project named A Bengali to English Word Aligner using Transformer Based Multilingual BERT. This project not only strengthened my understanding of NLP and Deep Learning concepts but also taught me the importance of iterative experimentation and strong communication skills while working on a team. My current research areas are Multilingual Word Alignments, Machine Translation, Generative AI, different LLM fine-tuning and Bengali Language tasks. In my free time, I love to write tech blogs and take nature photographs. I am excited to apply my skills and knowledge to real-world projects and contribute to a team of professionals in the field.
Education
Projects
A Donation Campaign Application using Flask Backend API and React Frontend where people can see the details of different donation campaigns and add them to a donation list. Utilise Local Storage and Pagination functionality to improve user experience.
Tools and Libraries: React, React-Router, Flask
Published a Python package in PyPI and Docker Image in Docker Hub which will make text preprocessing and working with text files much easier.
Made two GitHub Actions. One for Auto-Update-Changelog and the other for Auto-PyPI-and-GitHub-Release
Tools and Libraries: Python, PyPI, Poetry, Docker, GitHub Action CI/CD
Links: GitHub || Docker Image || GitHub Package || PyPI || Auto-Changelog || Auto-Release
A News Aggregator website that scraped different Bengali and English news from different news portals. It has an authentication system and a bookmarking system. It is also deployed using GitHub and Heroku.
Tools and Libraries: HTML, CSS, JavaScript, Python, Django, BeautifulSoup, Bootstrap, unittest, Sqlite, Render
Side Projects
Work Experience
Done Guided Research and Undergraduate Thesis under Bangladesh's one of the top NLP and Bangla NLP Researcher Dr. Nabeel Mohammed sir.
Thesis Topic:
A Bengali to English Word Aligner using Transformer Based Multilingual BERT
This is an NLP-based research project. Our fine-tuned model will help properly align a Bangla Word with its corresponding English word of a sentence. Thus we can transfer the feature of an English Sentence into a Bangla Sentence. After training with 2 Million Sentence Pairs, we got around 76% Accuracy.
Additional Research:
Implemented Bangla Parts of Speech Tagging, Named Entity Recognition, etc using our fine-tuned model.
Used Microsoft IL-POST Dataset to test our model. Preprocessed MS IL-POST Dataset to ensure proper Text Classification.
Created a 2000+ manually aligned Bangla to English Alignment Dataset to test our model.