Showrin Rahman
About
I am a Computer Science graduate from BRAC University (2025, CGPA 3.62/4.00), where I also developed an interdisciplinary perspective through a fully funded semester at Bard College, USA, under the OSUN Undergraduate Mobility Scholarship.
My research background spans machine learning, computer vision, and intelligent computing, with peer-reviewed publications and research work appearing through IEEE (EUROCON and ACDSA), Springer (IFIP AICT). Alongside my academic research, I have gained hands-on industry R&D experience developing AI-driven systems and applications.
My research interests center on applying artificial intelligence to real-world challenges, particularly through computer vision, machine learning, and intelligent systems. I am especially interested in developing practical, efficient, and deployable AI solutions with meaningful societal and environmental impact.
I am seeking graduate research opportunities with the goal of pursuing research at the intersection of AI and real-world applications.
Academic Result
BSc Computer Science
CGPA 3.62 · BRACU '25
Exchange Scholar
Bard College, USA
OSUN Mobility Scholar
Publications
Peer-Reviewed Papers
IEEE · Springer
Impact & Goal
AI for Real-World
Societal Impact
Education & Activities
BRAC University, Dhaka, Bangladesh
June 2021 – September 2025
BSc in Computer Science (CS)
CGPA: 3.62 / 4.00
Relevant Coursework:
Bard College, Annandale-On-Hudson, New York, USA
Jan 2024 – May 2024
Exchange Semester, Spring 2024
Received acceptance for the Spring '24 semester at Bard College, New York, USA through the OSUN Undergraduate Mobility Programme for sustainability and climate. Besides academic exploration, working on my project, CrisisCrew, focusing on disaster resilience was part of the journey.
Holy Cross College, Dhaka, Bangladesh
2018 – 2020
Higher Secondary Certificate
GPA: 5.00
Co-Curricular Activities
Bracu Mongol Tori, Senior I
Feb 2022 – June 2024
- Developed an offline rover mapping system with 20% efficiency gain, worked with OpenCV and YOLO v5, and built the BRACU Mongol Tori website frontend.
Robotics Club of BRAC University — ROBU, Secretary
Nov 2022 – June 2024
- Served as a panelist in a 40-member Research and Project Management department, managed the flagship Basics of Robotics course, and developed 5+ projects using ROS, OpenCV, Arduino, and Raspberry Pi.
Publications
Masked Autoencoder and Mamba-Based Self-Supervised Segmentation of SAR Imagery for Riverbank Erosion Detection
Showrin Rahman — Undergraduate Thesis, BRAC University, 2025
Develops a self-supervised framework combining Masked Autoencoders (MAE) and Mamba State-Space Models (SSMs) for semantic segmentation under sparse supervision. Integrates GAN-based temporal reconstruction to resolve missing time-series observations, automating land-water boundary delineation without dense manual labels.
Multi-Scale Feature Fusion with Adaptive Attention for Robust Image Deblurring
S. Rahman, S. H. Rahman, K. Kim, M. Jawad — IEEE EUROCON 2025, Poland
Proposes MSFAA with the CAFI cross-attention module for spatially-varying blind image deblurring. Demonstrates +2.1 dB PSNR, +0.043 SSIM, and 18% lower compute vs. SOTA baselines. Successfully deployed into production OCR pipelines for degraded business document recognition.
TWIG: Two-Step Image Generation Using Segmentation Masks to Prevent Copyright Infringement
Showrin Rahman et al. — IFIP AICT (Vol. 705), Springer, 2024
Designs a two-stage diffusion pipeline where intermediate segmentation masks decouple structural geometry from texture, synthesizing high-utility training datasets while preventing reproduction of copyrighted visual features.
Riverbank Change Detection in the Jamuna River Basin, Bangladesh: A Remote Sensing and Machine Learning Approach
Showrin Rahman et al. — IEEE ACDSA, 2025
Develops a multi-temporal change-detection framework combining multi-spectral satellite imagery, spectral index engineering (NDWI, NDVI), and supervised ML classifiers in Google Earth Engine to model dynamic morphological boundaries at scale.
TextDiffuser-RL: Efficient and Robust Text Layout Optimization for High-Fidelity Text-to-Image Synthesis
Showrin Rahman et al. — arXiv:2505.19291, 2025
Introduces reinforcement-learning-guided character-level layout optimization for text-to-image diffusion models, significantly improving glyph rendering accuracy, character spacing, and spatial visual coherence over baselines.
Research & ML Projects
Implementation and experimental work underlying or adjacent to publications.
FlowPay — AI Solution for Automatic Document Recognition
276 Holdings Co., Ltd. — AI & Data Division • Seoul, South Korea •
Jul 2025 – Present
Role: ML & DL Engineer, AI Researcher (Independent Researcher)
Developed OCR pipelines using EasyOCR and Tesseract for Automatic Document Recognition for FlowPay. Applied document-specific image preprocessing and MSFAA-based deblurring to improve OCR accuracy on degraded business documents. R&D area: Computer Vision and Intelligent Document Processing (IDP).
Certificate of R&D Collaboration issued Aug 2025 · 276 Holdings Co., Ltd.
Multi-Spectral Surface Pattern Classification
Computer Vision · Deep LearningCNN-based multi-class classifier on 13-band EuroSAT multispectral imagery to detect land-cover patterns and ecological variations at scale.
GitHub →
CrisisCrew — Fire Safety & Disaster Management Platform
A tech-driven social entrepreneurship solution addressing fire safety and disaster management in Bangladesh. CrisisCrew integrates IoT, GIS, and AI to empower trained community responders for faster, smarter fire hazard response — covering Preparedness, Response, Recovery, and Mitigation.
GitHub →
BiLSTM Part-of-Speech Tagger
NLP · Sequence ModelingBidirectional LSTM achieves 97% accuracy on token-level POS tagging (POS 1 dataset). Demonstrates sequence-to-sequence neural labeling.
GitHub →
MangoScan — Fine-Grained Image Classifier
Computer Vision · PyTorchFine-grained classifier for 10 Bangladeshi mango varieties using PyTorch transfer learning — 95.05% accuracy. Streamlit demo.
GitHub →
CrisisCrewBot — AI Emergency Response System
LLM Application · ReactAI-powered fire emergency data collection system for Dhaka, integrating Gemini AI and real-time Google Sheets reporting.
ThesisConnect — Academic Collaboration Platform
Full Stack · Research ToolsEcosystem for thesis management, research sharing, and scholarly communication. Built with React, Node.js, Express, MongoDB.
LLM Resource Tracker
Browser Extension · Sustainable AIPrivacy-first browser extension estimating token, energy, and water consumption of LLM sessions on-device. Supports ChatGPT, Claude, Gemini.
GitHub →Industry Experience
AI Application Developer — ACI PLC, Dhaka
Feb 2026 – Present
- Designed a 9-stage multi-agent generative pipeline (StyleGenAI) that enforces cultural and structural constraints on garment generation through a rule-based verification layer — a hybrid symbolic–neural approach to controllable generation.
- Built an LLM-based automated assessment system (AnyExamAI) for open-ended response generation and scoring, addressing evaluation bottlenecks in large-scale language testing.
- Currently investigating diffusion-based virtual try-on, focusing on garment-consistent image synthesis under pose and identity constraints.
- Developed an agentic document and scheduling assistant (Agent MD) integrating Google Drive and Calendar, enabling natural-language interaction over documents and schedules.
Associate AI Engineer (previously AI Intern) — Ontik Technology, Dhaka
July 2025 – Feb 2026
- Designed a retrieval-augmented conversational system (EzeeChat) using vector-based semantic search (pgvector) over unstructured institutional data, deployed for the University of Toronto.
- Built an agentic data-intelligence pipeline (RearWindow) combining automated ETL with LLM agents for spatial pattern analysis and reporting.
- Developed a multi-agent reasoning system (Properta) coupling LLM-based inference with structured data retrieval for property valuation and market analysis.
Research Collaborator (R&D) — 276 Holdings Co., Ltd.
AI & Data Division · Seoul, South Korea (Remote)
July 2025 – Dec 2025
- Contributed to an industry R&D project in computer vision and intelligent document processing (IDP), developing OCR pipelines with EasyOCR and Tesseract using document-specific preprocessing.
- Improved document recognition accuracy through MSFAA-based image deblurring, extending research published at IEEE Eurocon 2025, and conducted systematic experimental evaluation and prototyping.
Coding Instructor (Part-Time, Remote) — Dreamers Academy, Dhaka
Dec 2024 – Jul 2025
- Taught 100+ students beginner-level Python, JavaScript, HTML, and CSS.
Quality Assurance Analyst — Sysonex Development Limited, Dhaka
Jun 2024 – Dec 2024 · Front-end & Design Department
- Conducted QA processes for front-end and design deliverables; tracked issues and workflows with Jira and Git.
- Awarded Certificate of Completion (Dec 2024) recognizing dedication and contributions to Sysonex.
Technical Skills
Research & ML
Languages & Tools
Language Proficiency
Awards & Honors
Scholarships, competitions, and recognition
OSUN Undergraduate Mobility Scholarship
Spring 2024 · Bard College, New York, USA
Fully funded scholarship covering tuition, accommodation, living expenses, insurance, and travel — awarded on academic merit and research potential.
2nd Runner-Up — 3rd Kibo Robot Programming Challenge
2022 · Japan Aerospace Exploration Agency (JAXA)
Competed in JAXA's international robotics programming challenge using the ISS onboard computer.
Top 5 Finalist — FOREWARN Disaster Hackathon 2.0
Among 84 teams
Developed an AI-driven disaster preparedness solution and placed in the top 5 out of 84 competing teams.
Top 15 — Banglalink Womentor Program
Among 400 applicants
Selected in the top 15 out of 400 applicants for a mentorship program supporting women in technology.
University Innovation Hub Program (UIHP) Selection
BRAC IC4
CrisisCrew — our disaster preparedness initiative — was selected for UIHP as part of the BRAC IC4 innovation ecosystem.
Hult Prize Semi-Finalist — Team FLASH NEXUS
BRAC University
Team FLASH NEXUS advanced to the Semi-Final Round of the Hult Prize at BRAC University for a social entrepreneurship venture.
Media & Appearances
Interviews, talks, and film appearances
Tech Talk EP 06 — Generative AI
Ontik Technology · AI Intern Interview
"AI is unpredictable and that is exactly what makes it so exciting." — Showrin shares her thoughts on how Generative AI is reshaping the way we think about technology.
Watch on YouTube →
IEEE ACDSA 2025 — Research Presentation
Conference Talk · Peer-Reviewed
Presenting "Riverbank Change Detection in the Jamuna River Basin, Bangladesh: A Remote Sensing and Machine Learning Approach" at IEEE ACDSA 2025.
Watch on YouTube →
TMI Project: Bard (2024)
Documentary · Bard College, New York
Honored to have spoken at the TMI Project Bard 2024 cohort. Sharing my journey, including the struggles and strengths, was an incredibly rewarding experience. Grateful for the opportunity to connect with such inspiring individuals.
Watch on YouTube →
48-Hour Film Challenge — Bard College
Short Film · Spring 2024
Participated in a 48-hour film challenge at Bard College — exploring storytelling from in front of the camera, working with a diverse team under tight time constraints.
Watch on YouTube →