Machine Learning, Biomedical Image & Signal Processing Enthusiast | Python Programmer
Current Time: Loading...
My career began with curiosity—about how patterns hide inside images, how meaning is buried in text, and how biological signals encode the state of the human body. As an electronics engineer, I learned how systems behave. As a machine learning engineer, I learned how to extract insight from them. Today, I work across computer vision, NLP, and signal processing, with my core specialization in biomedical AI and building models that are transparent, reliable, and clinically meaningful.
My work in computer vision spans natural images and advanced medical scans. I have built and deployed models for disease detection in X-rays, MRIs, and CT scans, segmentation of anatomical regions, and precision enhancement of blood-cell imagery using generative models. What I focus on most is interpretability. Using methods like Grad-CAM, multi-branch CNNs, and hybrid transformer architectures, I design systems that highlight the diagnostic patterns experts rely on. This approach guides every imaging project I work on, ensuring models deliver clarity as well as accuracy.
Biomedical signals are some of the most information-rich forms of data. I have developed deep learning architectures for ECG and EEG interpretation, including EXGNet for arrhythmia classification and CARDxnosis for automated ECG-driven diagnostic reporting. My work includes designing custom loss functions, creating temporal attention mechanisms, and aligning model behavior with clinical heuristics such as ST-segment shifts and QRS width changes. My goal is always to produce systems that extract clinically meaningful patterns and present them in ways that support real medical decision-making.
Some biological systems cannot be understood through images alone. While working with fMRI-based brain connectomes, I shifted from traditional CNNs to graph neural networks. I built models that treat each brain as a connectivity map and classify conditions such as ADHD and ASD based on variations in neural pathways. This work blends computational neuroscience with deep learning, producing insights that extend beyond visual patterns and into the domain of structural and functional relationships.
Beyond images and signals, I also work with natural language processing. I build systems that understand and generate structured information, including toxic text classifiers, clinical report generators, domain-specific summarizers, and transformer-based medical text interpreters. This work allows me to integrate model predictions with readable documentation, improving communication between AI systems and human users.
My work extends beyond research. I design end-to-end systems that move from experimentation to deployment. This includes optimizing neural networks for edge devices like Raspberry Pi, developing Flask and FastAPI infrastructures for hospitals, collaborating with clinicians in Dhaka, Delhi, and Tokyo, and building tools that integrate seamlessly into existing workflows. For me, impact comes from creating solutions that function reliably in real environments.
I work across computer vision, NLP, and biomedical signal processing, but biomedical AI is where my experience has the deepest roots. This field demands transparency, reliability, and trust. My goal is to convert complex data into systems that support real decision-making—models that identify subtle patterns, provide clear reasoning, and integrate naturally into clinical workflows.
If your work requires strong engineering, deep learning expertise, and an understanding of biomedical complexity, I’m always open to collaborate.
Teton Private Ltd., Dhaka, Bangladesh
July 2023 - Present
Driving innovation in R&D by developing advanced machine learning models for biomedical applications and beyond.