Shoumik Sharar Chowdhury
ML Engineer specializing in computer vision and MLOps at Andium. Passionate about data for good and solving complex machine learning problems.
Research
My research experience spans both robotics and computational social science, focusing on practical applications of machine learning and data analysis.
Research Areas
Robotics & Computer Vision
As part of my undergraduate thesis under Dr. Tarem Ahmed, I developed a low-cost robot capable of interior mapping using sonar-based SLAM (Simultaneous Localization and Mapping). This work was presented at TENCON 2018, one of IEEE Region 10’s leading technical conferences.
Computational Social Science
At KolpoKoushol, a knowledge initiative led by Dr. Nazmus Saquib, I worked on a project analyzing political discourse patterns in Bangladeshi news media over a decade. Our research revealed how cohesive groups of powerful political leaders form through news coverage, and how these elite networks evolve as news cycles change over time. This work was published at NeurIPS 2018 in the ML4D workshop.
Publications
Statement Networks: A Power Structure Narrative as Depicted by Newspapers
Shoumik Sharar Chowdhury, Nazmus Saquib, Niamat Zawad, Manash Kumar Mandal, Syed Haque
ML for the Developing World (ML4D) Workshop at NeurIPS 2018
Abstract:
This paper examines political speech coverage across three major English-language newspapers in Bangladesh. Our analysis reveals a tightly connected group of political leaders who dominate the discourse, often focusing on themselves and each other while overlooking broader national concerns.
Sonar-based SLAM Using Occupancy Grid Mapping and Dead Reckoning
Muhammad S. A. Khan, Shoumik Sharar Chowdhury, Nafis Niloy, Fatema Tuz Zohra Aurin, Tarem Ahmed
IEEE Region 10 Conference (TENCON 2018)
Abstract:
We propose an affordable method for indoor mapping using sonar sensors, leveraging occupancy grid mapping and dead reckoning to enable autonomous navigation for mobile robots.