Advancing AI with explainable rigor and precision.
I am S. M. Mohiuddin Khan Shiam, a Computer Science graduate and researcher specializing in Explainable AI (XAI), Computer Vision, and High-Frequency Time Series. Author of peer-reviewed Q1 research in Elsevier Array, creator of Nano3D, and active contributor to 11+ global open-source foundations including Microsoft, Scikit-learn, and Kubernetes.
Q1
Elsevier Published
486K
Nano3D Params
11+
Global OSS Orgs
17+
Core Repositories
Academic & Research Identifiers
Verified scientific registries, citation databases, open-source repositories, and developer hubs.
Passionate about engineering intelligent, explainable systems.
I am a Computer Science Graduate from BRAC University and a Ph.D. Aspirant with strong academic grounding in software systems, artificial intelligence, and peer-reviewed machine learning research.
My work focuses on bridging foundational mathematical theory with real-time, resource-constrained edge computing. Whether formulating ultra-lightweight 3D convolutional models for edge surveillance or building Bayesian-tuned hybrid ensemble pipelines for high-frequency macroeconomic forecasting, my goal is to produce transparent, auditable, and resilient machine intelligence.
Beyond academic research, I actively contribute to world-renowned open-source projects powering millions of software engineers, and have served as Treasurer of the IEEE BRAC University Student Branch, fostering scientific leadership and technical workshops.
Computer Vision & Spatiotemporal Models
Designing ultra-compact 3D-CNNs (Nano3D) and ConvLSTM architectures for real-time video anomaly detection, spatiotemporal action recognition, and edge deployment across canonical surveillance benchmarks.
Explainable Artificial Intelligence (XAI)
Integrating post-hoc and intrinsic interpretability frameworks (Grad-CAM, SHAP, LIME, Partial Dependence Plots) into deep networks to generate verifiable, policy-grade attribution maps.
Time Series & Hybrid Ensembles
Engineering multi-tier ensemble architectures combining gradient-boosted trees with Bayesian optimization for robust high-frequency macro-financial forecasting (HybridStack, test RMSE 0.0006).
Quantum ML & Medical Diagnostics
Investigating Parameterized Quantum Circuits (PQCs) and Variational Quantum Classifiers with Qiskit/PennyLane, alongside 3D Vision Transformers for early cancer detection and stroke lesion segmentation.
Peer-Reviewed Research & Thesis
Published Q1 journal articles, manuscripts under review, and academic thesis.
HybridStack: Explainable hybrid ensemble for accurate inflation forecasting with high-frequency macro-financial data
- Engineered a multi-tier hybrid ensemble combining gradient-boosted decision trees (XGBoost, LightGBM, CatBoost) with regularized meta-regressors.
- Optimized hyperparameters via rigorous Bayesian search across 10 macro-financial sources, yielding a held-out test RMSE of 0.0006.
- Integrated SHAP, LIME, and Partial Dependence Plots (PDP) to generate policy-grade interpretability of critical economic predictors.
Nano3D: An Explainable Ultra-Lightweight Deep Learning Model for Surveillance Video Anomaly Detection
- Formulated a resource-efficient 3D-CNN architecture containing only 486K parameters (1.86 MB model footprint) for edge-device video anomaly detection.
- Attained 88.7% accuracy, 0.947 AUC, and 7.19 ms inference latency across 5 canonical surveillance datasets, outperforming heavier models.
- Integrated Grad-CAM spatiotemporal saliency heatmaps to visually explain and verify automated anomaly localization.
Deep Learning Frameworks for Suspicious Activity Detection
- Benchmarked deep spatiotemporal learning architectures under uniform conditions across 5 canonical public surveillance datasets: UCF-Crime, XD-Violence, RWF-2000, UBI-Fights, and SCVD.
- Proposed Nano3D, an ultra-compact 3D convolutional model containing only 486K parameters, achieving 88.7% accuracy, 0.947 AUC, and 7.19 ms inference latency.
- Implemented Grad-CAM spatiotemporal saliency heatmaps to visually explain and attribute automated anomaly localization.
Active & Ongoing Research Frontiers
Quantum Machine Learning (QML)
Formulating hybrid quantum-classical algorithms, Parameterized Quantum Circuits (PQCs), and Variational Quantum Classifiers (VQC) utilizing Qiskit and PennyLane for high-dimensional constrained optimization problems.
AI in Medical Image Diagnostics
Designing 3D Vision Transformers (ViT) and CNNs for early cancer detection (mammography, histopathology) and acute ischemic stroke lesion segmentation from brain CT/MRI scans with Grad-CAM and SHAP interpretability.
Featured Technical Projects
Open-source software systems, deep learning models, and full-stack engineering.
Nano3D: Explainable Surveillance Anomaly Detection
An ultra-compact 3D convolutional neural network designed for edge-device anomaly detection across UCF-Crime, XD-Violence, RWF-2000, UBI-Fights, and SCVD datasets.
Deep Learning vs. Statistical Stock Forecasting
Empirical comparative time-series forecasting evaluating LSTM Recurrent Neural Networks against classical ARIMA models on HP Inc. historical equity data, analyzing volatility and stationarity.
SkillShikhi: Skill-Sharing & Learning Platform
A full-stack collaborative educational web application with MVC architecture, JWT authentication, user dashboards, responsive UI, dynamic content routing, and secure RESTful APIs.
Autonomous Line-Following Hardware Robot
An autonomous line-tracking hardware vehicle engineered for the Digital Logic Design curriculum using discrete logic gates, IR optical sensors, comparator ICs, and H-bridge motor drivers without microcontrollers.
Global Open Source Contributions
Active contributor to 11+ global foundations across 17+ mission-critical repositories powering millions of developers.
Enhanced IntelliSense, debugging performance, and extension API features.
Performance optimizations and advanced shell integration enhancements.
Cross-platform compatibility, kernel bridges, and filesystem interoperability.
Developer productivity utilities and UI/UX workflow optimizations.
Multi-agent conversation frameworks and agentic workflow orchestration.
Port management, dependency graph resolution, and build stability.
Command-line package installer improvements and manifest validation.
Core algorithm optimizations, documentation, and estimator consistency.
Model architecture layers, testing suites, and training pipeline enhancements.
YOLO object detection tuning, tracking stability, and deployment pipelines.
Neural network hardware acceleration, runtime profiling, and model conversion.
Controller stability, node management, and scheduler optimization.
Automated Kubernetes cluster provisioning and cloud provider integration.
DAG scheduling efficiency, provider operators, and task execution pipelines.
GitOps synchronization controllers and Helm release reconcilers.
Production-clean Python algorithm implementations and unit tests.
Mathematical visualization library, vector transforms, and rendering features.
Biomolecular structural modeling and 3D molecular interaction simulations.
Technical Skills & Stack
Programming languages, deep learning frameworks, AI paradigms, and cloud engineering.
Programming Languages
Machine Learning & DL
AI Paradigms & Methods
Web & Cloud Engineering
Experience & Education
Academic milestones, research appointments, and institutional leadership.
💼 Work & Leadership
Machine Learning Researcher
Conducting advanced AI research in high-frequency time-series forecasting, neural networks, and Bayesian-tuned hybrid ensemble models. Engineering end-to-end ML data pipelines and authoring peer-reviewed scientific articles published in international Q1 journals.
Treasurer
Managed student branch fiscal budgeting, financial tracking, funding allocations, and transparent institutional reporting. Co-organized major technical workshops, seminars, and training programs, including the flagship “Smart City & IoT” workshop.
🎓 Education
Bachelor of Science in Computer Science
Comprehensive coursework in Algorithms, Software Systems, Artificial Intelligence, Deep Learning, and Computer Architecture. Thesis on spatiotemporal deep learning for surveillance anomaly detection. Served as Treasurer of IEEE BRACU Student Branch (2024).
Higher Secondary Certificate (HSC) – Science Group
Rigorous background in Physics, Chemistry, and Higher Mathematics. Awarded Academic Excellence in National Board Examinations.
Secondary School Certificate (SSC) – Science Group
Strong foundation in Science and Mathematics. Active leader of the School Science Club and defender in the School Football Team.
Honors, Awards & Certifications
Competitive distinctions, athletic championships, and professional accreditations.
🏆 Honors & Awards
Top 10% Global Talent
Issued by The Talent Games • November 2024
RS-60 Cricket Cup Champion
BRAC University Residential Semester • July 2022
Academic Excellence in National Board Examinations
Higher Secondary Certificate (HSC), Science Group • 2020
📜 Professional Certifications
Enhance Your Leadership Skills
LinkedIn Learning • Issued Jan. 2025
Agile Foundations
International Institute of Business Analysis™ (IIBA®) • Issued Dec. 2024
Time Management Fundamentals
Project Management Institute (PMI) • Issued Aug. 2024
Communication Hacks
10 Minute School • Issued Jun. 2024
Professional Academic References
Esteemed professors, thesis supervisors, and research collaborators.
Dr. Md. Golam Rabiul Alam
School of Data & Sciences, BRAC University
Dr. Mahdy Rahman Chowdhury
North South University
Dr. Amitabha Chakrabarty
School of Data & Sciences, BRAC University
Md. Ariful Islam Sanim
University of South Carolina, USA
Articles & Writings
In-depth tutorials, deep learning walk-throughs, and software engineering reflections.