Education
Bachelors in Computer Science, Brac University (2020 – 2024)
Relevant Coursework:
- Data Structures
- Algorithms
- Artificial Intelligence
- Natural Language Processing
- Data Science
- Randomized Algorithms
Work Experience
Undergraduate Teaching Assistant/Student Tutor
Department of Mathematics and Natural Science, Brac University (2022 – 2024)
- Assigned Courses: STA201: Statistics and Probability, MAT120: Integral Calculus and Differential Equations.
- Graded assignments and quizzes for three sections each semester, totaling over 90 students, ensuring timely feedback to support academic improvement.
- Provided 15 hours of weekly consultations to students who needed assistance with their assignments and understanding class materials.
Junior Engineer, Control and Software
Bracu Mongol Tori, Brac University (2021)
- Developed a user interface using Python and Tkinter that interacts with environmental data obtained by Arduino sensors and plots their graph in real time.
- This project was part of the University Rover Challenge (2021).
Skills
Tools & Technologies
PyTorch, Langchain, Numpy, Pandas, Matplotlib, Scikit-Learn, NLTK, spaCy, Pygame, GitHub, Version Control, Linux, LaTeX, React JS (Basic)
Programming Languages
Python, C/C++, Java, JavaScript, HTML & CSS (Basic)
Projects
Chess Game
- Built a fully functioning Chess game using Python and Pygame.
- Implemented all major functionalities such as castling, pawn promotion, checkmates, stalemates, pawn en passant, and clock maneuvers.
Person Name Extractor for Bangla
- Fine-tuned a pre-trained Bangla-BERT Model and built a Named Entity Recognition (NER) system to extract names from Bangla texts.
- Used the Bangla subset from the WikiANN dataset which includes train, test, and validation sets with tokenized texts and named entity tags.
- The model achieved an F1 score of 0.97 across the validation set after training for 3 epochs.
Hallucination Reduction in Large Language Models using Chain of Verification
- Reproduced the proposed methodology of the paper titled "Chain of Verification Reduces Hallucination in Large Language Models".
- Implemented a custom parsing algorithm using regular grammar to extract relevant information from LLM-generated responses.
- Applied 4-bit quantization on Mistral 7b for improved computational efficiency.
- Implemented the CoVe chain on the Wiki-data question-answering task, increasing the micro-averaged precision from 0.12 to 0.25.
Pathfinding Visualizer
- Implemented Dijkstra and A* shortest pathfinding algorithms using Python.
- Using a 2D graphics module named Pygame, visualized the workflow of those algorithms in finding the optimal path.
Sorting Visualizer
- Implemented famous sorting algorithms such as Bubble Sort, Selection Sort, Insertion Sort, and Merge Sort in Python.
- Visualized their workflow using Pygame.
Wordle Game
- Recreated the famous word game "Wordle" in Python using Pygame.
- Implemented an API for the main game functionalities that can be used for building the user interface from scratch.
- Developed a custom algorithm for finding the solution of the puzzle within 5 tries.
Bracu CGPA Calculator
- Built a CGPA calculator web app using React-JS and JavaScript tailored toward Brac University's grading system.
- This app uses students’ previously completed credits, their GPA, and current semester results to calculate cumulative grading points.
Hate Speech and Offensive Language Detection
- Classified labeled tweets as ’hate speech’, ’offensive language’, and ’neither’ with high accuracy.
- Performed oversampling and text-data preprocessing.
- Applied three machine learning algorithms: Logistic Regression, Naive Bayes Classifier, and K-th Nearest Neighbor to achieve 96.78% accuracy.
Sudoku Solver Visualizer
- Built a Sudoku Game in Python that uses an external API to fetch a random unsolved game.
- Using Pygame, demonstrated the advancement process of a backtracking algorithm in solving the game.
Publications
S.M. Abrar Mustakin Taki, Showmick Kar, Soumik Deb Niloy, Mazharul Islam Rakib, Abdul Al Nahid Biswas. "Mitigation of hallucination and interpretations of self attention of Mistral 7B AI to analyze and visualize context understanding ability of large language models." Undergraduate Thesis, BRAC University, 2024. Thesis Advisor: Dr. Farig Yousuf Sadeque. Available at: http://hdl.handle.net/10361/22762
Maidul Islam, Mohammad Sadman Tahsin, Farjana Alam, Sadab Sifar Hossain, Arnob Deb, Showmick Kar. "Understanding Convolutional Neural Network’s behavior for Alzheimer’s disease on MRI." 2022 IEEE 13th Annual Information Technology, Electronics and Mobile Communication Conference (IEMCON). DOI: https://doi.org/10.1109/IEMCON56893.2022.9946529
Interests
Reading, Music, Chess, Language, Culture.