Hello! I am Shahrar. Welcome to my website.
I hope you will enjoy the journey.

Let's Start

Name: Md Shahrar Fatemi

Post: Lecturer (On Leave)

At: CSE,
School of Data & Sciences,
BRAC University

Skill

OpenCV 85%
MatLab 75%
AWS 50%
Python 90%
Other Coding Languages : C/C++, Java, HTML, CSS
Other Frameworks/Libraries : React, Node
Database Proficiencies : MongoDB, PostgreSQL, Firebase
About Me

I am a graduate from the Department of Computer Science and Engineering, Bangladesh University of Engineering and Technology (BUET). Currently, I am a PhD student in CS at Purdue University. In addition to my studies, I am working as a Graduate Teaching Assistant at Purdue University.

The vast field of Bioinformatics mesmerizes me. Currently, I am working under Dr. Md. Shamsuzzoha Bayzid on a project where we are trying to segment lung areas and their nodules in point cloud representation. We have already established an extensive structure for 3D visualization of lung areas in point clouds. The intersection of Biology, Medical Science, and Computer Science (especially Deep Learning, Image processing, and Imaging techniques) amuses me very much. During the whole 1.5 years of my remote research internship at XuLab, CMU, I have been involved in research projects on identifying cellular molecules from Cryo-ET images. Getting the Upsilon Pi Epsilon Award 2021 for academic excellence in the computing discipline is one of the best achievements of my life so far. I am honoured to be one of the other three winners from all over the world.

If I am asked to talk about the few good things in my life, I would say I have a very understanding family. I feel extremely lucky to be able to study in the topmost Engineering Institute of the counrty. My passion apart from studies, is playing football. I have played for my University in national levels.

During my undergraduate education, I worked under Professor Dr. A. B. M. Alim Al Islam in his project called DhakaSim - a Dhaka traffic management simulation system. I worked as a junior Research Engineer in Celloscope BD too, with a team motivated to address NLP problems in Bangla Language. Details on my research experience can be found in Research & Publications section.

Research & Publications

Areas of interest : Bioinformatics, Medical Images, Machine Learning,
Computer Vision and Ubiquitous Computing

Publications

Practical Analysis of Macromolecule Identity from Cryo-electron Tomography Images using Deep Learning

paper link
Conference: AIPR 2021

Co-Authors: Mostofa Rafid Uddin, Ajmain Yasar Ahmed, Kafi Khan, Xiangrui Zeng, Min Xu

It is essential to understand the cellular dynamics of macro-molecules to find out a remedy for diseases. Hence, we tried to identify cellular macromolecules using various methods involving both supervised and unsupervised ones. We analyzed the results of using them in Cryo-Electron Tomography images consisting of large macromolecules. Both methods have drawbacks, even though the unsupervised method DISCA showed promising results in identifying previously unseen macro-molecules.

SHONGLAP: A Large Bengali Open-Domain Dialogue Corpus

paper link
Conference: LREC 2022, ID : 885

Co-Authors: Syed Mostofa Monsur, Sakib Chowdhury and Shafayat Ahmed

Bangla is a low resource language. There is not enough data to classify human sentiments from them, let alone understand dialogues. Our system's pipeline could smoothly diarize human speeches in Bangla into utterances belonging to different human beings and annotate them. In addition, we added sentiment analysis to it. It's a vibrant (more than 7000 dialogues) resource of Bangla Conversation prepared by us.

Research Experiences

Identifying Genomic Islands

Supervised by : Dr. Md. Shamsuzzoha Bayzid
Status : running

SSG-LUGIA (a previously published paper from the lab) is a completely automated and unsupervised approach for identifying GIs and horizontally transferred genes. We are currently analyzing similar tools and bringing new features into it so that external information like protein annotations can be put into the features in order to determine genetic islands more precisely.

Lung Nodule Detection Analysis: Deep Learning and Point Cloud based approach

Supervised by : Dr. Md. Shamsuzzoha Bayzid
Status : running

Detecting malignant nodules in the lungs of a human body is a very tough task. Again it is vital to identify them as early as possible. Though most of them are benign, some are malignant and might develop into cancerous lesions that are too big. We studied many conventional methods of detecting and classifying lung nodules. Generating point cloud data of lungs and identifying nodule areas has been our point of interest here. Hence, we developed a specific algorithm to segment lung areas from chest CT scans. After using a couple of point cloud features-based methods to detect nodules, we are considering using deep learning-based large-scale point cloud segmentation techniques. Here's a video of the segmented lungs in point cloud.

Practical Analysis of Macromolecule Identity from Cryo- ET Images using Deep Learning

Led by : Xu Lab at CMU
Status : published

Cryo-Electron Tomography images consisting of large macromolecules help us to visualize the systematic 3D representation of cellular organelles inside single cells. We analyzed different processes of identifying molecular dynamics in those images. Supervised methods are very useful for correctly identifying macromolecules. But, it's too time-consuming to manually annotate each and every particle in a Cryo-ET image. Again, unknown particles won't be picked in those cases. That's when we come up with an unsupervised method DISCA (Deep Iterative Subtomogram Clustering Approach).

Unsupervised Object Detection in Cryo-ET Images

Led by : Mostofa Rafid Uddin
Status : running

We are figuring out particles of any type in datasets that are almost unannotated. As unsupervised methods have a higher tendency to result in a higher ratio of false positives, we are considering working in a semi-supervised manner.

DhakaSim

Supervised by : Dr. A. B. M. Alim Al Islam Status : partially finished.

Because of intrinsic differences among internal infrastructures, the traffic systems deployed in developed countries would almost certainly fail catastrophically in developing ones. In this extensive project, we are studying and building a model to simulate and analyze the behavior of Dhaka traffic in various real-life situations.

Projects

They are mostly related to my course works and extra-curricular activities.

Web-based Software

SoccerBoard

This one is a demo software to maintain training and game schedules for the team members, squad management platform for the coach, interactive whiteboard for the training plan demonstration etc.

Micro-Controller

ATmega FireFigher

We created a prototype of an automated and primary fire prevention system. The details demonstration on this will be found here.

Android App

Quizzy

In this app,one can take various types of quizzes and have personal evaluation(built during pandemic).

Network Security

TCP Session Hijacking

In TCP session, the host at first logs into the server with some credentials and then they do a three way handshake based authentication. The attacker tries to break into the session after this. I made this project as part of my Network Security Course.

Machine Learning

Climate Predictor

As part of our Machine Learning course, I had to prepare a simulation of Baum Welch Learning (a special EM method) using a Hidden Markov Model, the given scenario belonged to two different climate patterns. Given previous data, our task was to approximate parametes of the Markov Model.

Network & OOP

8 Ball Pool

As part of our Object Oriented Programming course, we prepared a replication of the original 8 ball pool game, which can be played by two players connected by LAN Network.

Awards & Achievements

Global Award


Upsilon Pi Epsilon Award 2021


The Executive Council of Upsilon Pi Epsilon (UPE) wishes to encourage academic excellence for students in the computing discipline. UPE initiated this award with the IEEE Computer Society, as the largest association of computer professionals, to raise the importance of academic achievement in our future computer professionals. I was blessed to be one of the global winners of the award.


National & Educationals

Education

CGPA: 3.91 / 4.00
Year: 2022

B. Sc. In Computer Science & Engineering

GPA : 5.00 / 5.00
Year: 2016

Higher Secondary Certificate

GPA : 5.00 / 5.00
Year: 2014

Secondary School Certificate

Experiences

Dept. of CSE, BRAC University

Lecturer (On Leave)
1 Jun. 2022 - present

Courses taught:

  • CSE221 : Algorithms
  • CSE230 : Discrete Mathematics

Dept. of CS, Purdue University

Graduate Teaching Assistant
14 Aug. 2023 - present

Courses teaching:

  • CSE57900 : Bioinformatics Algorithms

Dept. of CSE, Bangladesh University of Engineering and Technology (BUET)

Lecturer (Adjunct)
16 Nov. 2022 - 30 Apr. 2023

Courses taught:

  • CSE311 : Data Communications
  • CSE204 : Data Structures and Algorithms - 1 : Sessional
  • CSE102 : Structured Programming Language : Sessional

XuLab at CMU

Remote Research Intern (Voluntary)
1 Jun. 2021 - 31 Dec. 2022

Team Leader: Mostofa Rafid Uddin

  • Current Project: Analysis of Unsupervised Object Detection Methods in Cryo ET Images
  • Previous Project: Practical Analysis of Macromolecule Identity from Cryo-Electron Tomography Images using Deep Learning

Celloscope BD

Research Engineer (Permanent)
1 May. 2022 - 31 May. 2022
Research Engineer (Part-time)
1 Dec. 2021 - 30 Apr. 2022