Hello!
I'm Shahariar Rabby, a CS PhD & Machine Learning Engineer who builds and deploys production-ready computer vision systems. From 3D medical imaging to OCR, I turn complex research into scalable ML solutions.
Get in touch shahariarrabby[at]gmail.com
I am a Computer Science PhD Candidate at UAB and a Graduate Researcher at the CLAIR lab. My work focuses on building deep learning models that solve high-stakes problems, from achieving 91% AUC in COPD prediction to enhancing segmentation in medical and structural imaging.
Before my PhD, I was a Machine Learning Engineer at Apurba Technologies, where I was promoted to Team Lead and boosted OCR accuracy by 5%. I have hands-on experience in the full ML lifecycle, from model development (PyTorch, TensorFlow) to deployment (Docker, FastAPI, AWS).
I am currently seeking a full-time Machine Learning Engineer or Applied Scientist role.
- Python
- C++
- C
- SQL
- JavaScript
- HTML/CSS
- PyTorch
- TensorFlow
- OpenCV
- CNNs
- Segmentation
- Docker
- FastAPI
- Flask
- AWS (EC2, S3)
- Triton
- REST API
- Git
- GitHub
- Linux
- Bash
- LaTeX
• Achieved 91% accuracy for COPD prediction on 3D CT scans for an NIH-funded project
• Collaborated on an NSF-funded project to improve the security of reading hand-marked ballots
• Delivered end-to-end ML projects that boosted OCR accuracy by 5%
• Mentored junior engineers and maintained high standards for ML software quality
Developed a state-of-the-art deep learning model that achieves 91% AUC on standard scans and 86% on low-dose scans, with an average prediction time of 27.8 seconds. The model uses advanced CNN architectures and data augmentation techniques to improve diagnostic accuracy for chronic obstructive pulmonary disease from 3D CT images.
Built a complete self-driving car system using computer vision, deep learning, and control theory. Implemented traffic light detection, lane finding, and vehicle control algorithms. The system successfully navigates through complex traffic scenarios with real-time decision making.
Created and published the largest dataset for Bangla characters, containing 673K instances. This dataset enhanced multiple OCR projects, achieving accuracy improvements of 18%. The dataset includes diverse demographics and has been widely adopted by the research community.
When I'm not in front of a computer screen, I'm probably reading books, watching movies, or crossing off another item on my bucket list. I am always happy to talk about research, outdoor activities, fieldwork, STEM equity, or science communication. Whether looking for a research collaborator or movie partner, do not hesitate to email me and/or follow me on social media.