Shahariar Rabby — Research
CS PhD, UAB. Dissertation: Precise Segmentation of Thin, Elongated, and Underrepresented Structures. CS Outstanding PhD Student Award, 2026. Advisor: Dr. Chengcui Zhang. 45+ publications, 1,000+ citations, h-index 17.
Areas
Applied AI & Analytics
Decision Support
Machine learning for decision support, long-tail learning, model evaluation and calibration.
Deployable AI (MLOps)
Systems
Scalable model serving, GPU inference optimization, latency-aware deployment.
Vision & Document AI
Computer Vision
Medical imaging, infrastructure monitoring, OCR/document pipelines, thin structure segmentation.
Responsible AI
High-Stakes Domains
Privacy-aware clinical/enterprise AI, human-in-the-loop validation, robust external evaluation.
Publications
Light Convolutional Neural Network for Detecting COPD: Multicenter Model Development and External Validation
medRxiv, 2025 (under review at The Lancet Digital Health)
STRIPE: Directional Feature Extraction for Pulmonary Fissure Segmentation
Manuscript in preparation
THREAD: Efficient Topology-Preserving Network for Thin Structure Segmentation
Manuscript in preparation
Enhancement of Bengali OCR by Specialized Models and Advanced Techniques
IEEE/CVF WACV Workshops, 2024
Gold Standard Bangla OCR Dataset: Data Preprocessing and Annotation
EMNLP Industry Track, 2023
Grants
NIH/NHLBI R01 HL151421
"Structural Determinants of Disease Progression in COPD" — PIs: Surya Bhatt, MD & Arie Nakhmani, PhD
NSF SaTC #2154589
"Collaborative Research: Bubble Aid" — PI: Chengcui Zhang | $397,302
Teaching & Mentoring
- Lecturer, Daffodil International University (2019–2020) — CSE 412 (AI), CSE 413L (AI Lab)
- Mentored 13 undergrads to first-author publications; multiple mentees pursuing PhDs
- Research Supervisor, DIU NLP & ML Lab (2019–Present)
- Workshop Organizer, WVLL @ IEEE/CVF WACV 2024
- Lead Organizer, WVLL 2026, University of North Texas
Service
- Webmaster, UAB Graduate Student Government (2024–2025)
- Peer Review: IEEE TDSC (9), IJMDEM (6), IEEE TMM (3), ACM Multimedia (6), WACV (4), ICMLA (3), AAAI (1), CVPR (1)
Artifacts
- Ekush — Bangla handwritten characters, 673K instances, 242 classes
- dCrack61K — Crack segmentation benchmark, 61,943 images, 36 sources
- ISHARA-LIPI — Bangla Sign Language, 1,800 images, 36 classes
References
- Dr. Chengcui Zhang — Dissertation Advisor, Professor of CS, UAB. [email protected]
- Dr. Sandeep Bodduluri — Asst Prof, Pulmonary, Allergy, and Critical Care, UAB. [email protected]
- Dr. Fuad Rahman — Founder & CEO, Apurba Technologies. [email protected]