Academic

Ph.D. in Computer Science

Institute Name: University of North Carolina at Charlotte(UNCC)

Research Topics: Distributed System and Parallel Computing

Start Date: August 2017

Estimated Graduation Date: May 2022

Bachelor of Science(B.Sc.)

Institute Name: Bangladesh University of Engineering and Technology (BUET)

Department: Computer Science & Engineering (CSE)

Passing Year: 2013

Higher School Certificate(HSC)

Government Science College (GSC)

Division:- Science

Academic Year:- 2005-2007

Secondary School Certificate(SSC)

Institute Name: Joypara Pilot High School

Division: Science

Passing Year: 2005

Research Experience

Institute: University of North Carolina at Charlotte(UNCC)

Position: Research Assistant(RA)

Duration: August 15th, 2017 - To Date

Responsibility: Work as a research assistant of Dr. Erik Saule, who is an associate professor at the University of North Carolina at Charlotte.

I like to work on parallel computing and distributed systems. We are interested in the buildup performance model for different graph problems. We are also interested in computer vector architecture.

• Vectorized Parallel Louvain Method

  Louvain method is a greedy optimization algorithm that maximizes the modularity of a partition of the network. Lots of different implementations exist for this method but none of them explore the advantage of the intel vector architecture. In this project we buildup a vectorized parallel Louvain method with the help of the intel Intrinsic programming.

  In our project, we explored the Intel Skylake and knight landing(KNL) processor. Each of them has 512 bits register, which means for a single precession floating point, it can provide vector length 16. So, it can perform 16 single-floating operations at a time.

• Performance Model for the Sparse Matrix-Vector Multiplication(SpMV)

  The main aim of this project is to buildup a performance model that can give approximate execution time and predict better solutions for graph analysis like Sparse Matrix-Vector Multiplication(SpMV) kernel. We dived our project into three different parts,

  1. Parallel Shared-Memory SpMV
  2. Distributed SpMV on CPU architecture
  3. Distributed SpMV on CPU GPU architecture

  We used the MPI standard to implement the SpMV on the distributed system and OpenMP to handle the parallel shared-memory SpMV. To perform SpMV on a distributed system, we partitioned the matrix into multiple sub-matrix such that each MPI process can perform SpMV on each individual sub-matrix. In our experiment, we used two different approaches to partition the matrix,

  • 1-D partitioning: METIS graph partition
  • 2-D partition: Random partitioning

Publications

• [Conference Paper] Md Maruf Hossain and Erik Saule. Postmortem Computation of Pageranl on Temporal Graphs. In 51st International Conference on Parallel Processing, pages 1–11, 2022. [pdf]




• [Workshop Paper] Md Maruf Hossain and Erik Saule. Impact of avx-512 instructions on graph partitioning problems. In 50th International Conference on Parallel Processing Workshop, pages 1–9, 2021. [link]




• [Poster] Md Maruf Hossain and Erik Saule. Postmortem graph analysis on the temporal graph. 2021.[link, poster]




• [Under Review] Md Maruf Hossain and Erik Saule. Performance Model of Iterated Solvers for the Distributed System.

Jobs

• Employer: Dynamic Solution Innovators

  Position: Software Engineer

  Duration: March 1st, 2014 - July 31st, 2017

  Contributions:

  Projects Name	Description	Technology Used
  Corporate Portal	The main purpose of the system is to maintain training process of corporate people from particular institute. Different Packages will be available, so they can buy particular opportunities and use them properly. Subscribed institute can customize the feature as they like.	Custom jxntm framework introduced by jenzabar company based on java, hibernate, ejb. Database Support: MySQL, Microsoft SQL Server, oracle. Front end: Angular js, Bootstrap, javascript, jquery, CSS3, html5 and support all modern browser. Server: jboss Server.
  Higher reach	Educational management system. It has two major Panel. Admin: desktop based admin panel which maintain all section, offering, payment system and also maintain affiliated organizations. Quick Admit: Web based this panel mainly used by student so, that they can submit registration and order different products.	Custom jxntm framework introduced by jenzabar company based on java, hibernate, ejb. Database Support: MySQL, Microsoft SQL Server, oracle. Front end: Extension js, Bootstrap, javascript, jquery, CSS3, html5 and support all modern browser. Server: jboss Server run in EC2 container. Use Amazon Web Services(aws): S3 Bucket, EC2 Container, SNS, SQS, Elastic Beanstalk, Cloud Watch, API gateway etc..
  Hiring App	The objective of this web based system is to advance recruitment process for different institute. People can apply to a institute and can able to take part in written test through this system. Interviewer can evaluate candidate on this portal.	Spring framework, hiberbate. Database Support: MySQL. Front end: Angular js, java script, jquery, CSS3, html5 and support all modern browser.
  OnlineApp	This system introduced for handle graduate students. It has two panel. One, student login system so, they can apply different session and course. Second, commander login panel so, teacher can recommend different student.	Custom jxntm framework introduced by jenzabar company based on java, hibernate, ejb. Database Support: MySQL, Microsoft SQL Server, oracle. Front end: Backbone js, Bootstrap, javascript, jquery, CSS3, html5 and support all modern browser. Server: jboss Server.

  
  

• Employer: Nazdaq Technologies

  Position: Junior Software Engineer

  Duration: February 17th, 2013 - February 28th, 2014

  Contributions:

  Project Name	Description	Technology Used
  Hospital Management	Developed Hospital Management System. Integrating different electronic medical instrument with core project. Automated medical billing services.	Framework: Spring Front end: Extension js, jquery, javascript, CSS3 Database: MySQL. Server: Apache Tomcat server.
  Banking Solution	Developed core banking and SMS banking solution. Implement annual Anti Money Laundering report for bank.	Framework: Spring Database: Microsoft SQL Server. Server: Apache Tomcat server.

  
  

Academic Projects

• Net-Scanner

  We implemented a network scanner to find all the shared folder in the same network. There is no need to remember the partner's network address, one can put the IP address range to explore.

• Minesweeper

  It is a well-known computer game, I implemented it by myself as a course project by using OpenGL. The game is initialized by a square grid with some hidden mine. Use can click any of the cells on the grid and need to explore the full grid without hitting the mine. If anyone clicks on the mine then the game is over and the user is lost the game. The main challenge was to recursively unpack all the cell that does not have any neighbors that contain mines.

Technical Skills

• Languages: C/C++, Java, Python, JavaScript, Jquery, 80x86 Assembly, Prolog, HTML, XML, CSS
• Cloud Computing: Hadoop, Spark, Amazon Web Services(AWS)
• Framework: .NET, Spring Framework
• Database: Oracle, Microsoft SQL Server, MySQL, SQLite
• Simulators: PSpice, Nachos, Matlab
• Design Tools: Visual Studio, IntelliJ IDEA, NetBeans, Eclipse, Photoshop, ER-Diagram, Spring Tool Suite
• Operating System: Windows, Android and other Linux distributions
• Hardware: AVR Programming for ATMEL
• Others PL/SQL, Lex, Yacc, UML, OpenGL, Shell Script, LATEX

Undergraduate Thesis Works

Title of thesis: Methods of Community Detection in Social Networks [pdf]

Supervisor: Prof. Dr. Abu Sayed Md. Latiful Hoque

Research Area: Graph Data Mining, Social Network Analysis

Our Contributions

• A detailed study on user interaction based community detection in social networks.
• We proposed a novel approach that calculates the cosine-similarity on the communication text among different actors.
• Rather than rely on the graph structure, our solution provides the community in the social network based on user interactions.

Contact Information