Bridging Theoretical AI & Production Systems
A multi-disciplinary trajectory spanning foundational machine learning research, full-stack software architecture, and scientific writing.
Md Tawhidul Islam
Graduate Researcher · Developer · Designer
Welcome to my corner of the digital world. I’m Tawhidul Islam - an AI/ML & Data enthusiast, technology explorer, and creative problem-solver. Here, I share my work, ideas, experiments, and the things I’m learning along the way. Explore, discover, and let’s build something meaningful with technology.
Academic & Code Links
Research Focus & Academic Narrative
I am a computer science researcher and software engineer operating at the convergence of mathematical machine learning, distributed infrastructure, and intuitive interface design. My graduate work concentrates on overcoming compute and memory constraints in long-context neural sequence processing.
My recent investigations explore structured state-space sequence models (SSMs) and hybrid recurrent formulations as viable sub-quadratic alternatives to dense multi-head attention. By formulating hardware-aligned recurrent operators and custom fused GPU kernels, our research seeks to expand context horizons without the severe memory bottlenecks that impede foundation models.
Outside the laboratory, I design production full-stack systems with obsessive attention to user experience, type safety, and clean architecture. I believe rigorous technical writing is a direct expression of clear thought; each dispatch explores theoretical concepts with uncompromising mathematical and practical clarity.
6+ Yrs
Engineering
3.96
Graduate GPA
100%
Open Code
Zero-Spam
Publications
Jump to Portfolio Segment
Core Research Pillars & Working Principles
Theoretical frameworks guiding laboratory investigations alongside practical engineering heuristics.
Efficient Deep Learning
Designing hardware-aligned recurrent architectures (SSMs, Mamba) and fused GPU kernels that diminish the quadratic compute bottleneck of traditional attention mechanisms.
Mechanistic Interpretability
Probing latent representations and attention circuits in multimodal foundation models to convert black-box predictors into causally verified conceptual graphs.
Systems & Interface Design
Translating intricate mathematical and distributed abstractions into responsive, accessible, high-contrast digital interfaces that facilitate scientific exploration.
Methodological Rigor
Insisting on reproducible open benchmarks, publicly shared artifacts, mathematical proofs of convergence, and zero-compromise architectural standards.
First-Principles Understanding
Never accept a tool, abstraction, or loss function without understanding its underlying linear algebra, hardware memory hierarchy, and algorithmic complexity.
Bypass Unnecessary Abstractions
When high performance and clean architecture matter, rely on typed raw SQL pools, zero-overhead CSS custom property tokens, and custom fused compute kernels.
Open Scientific Dissemination
Research that cannot be independently audited or reproduced does not advance science. Every publication should be accompanied by verified code repositories and weights.
Design as an Intellectual Tool
Visual precision, restrained typography, and intentional micro-interactions are not decorative; they are cognitive instruments that clarify complex thought.
Technical Skills & Architecture Stack
A categorized breakdown of machine learning frameworks, full-stack systems, design token tooling, and research environments.
Data Science and AI/ML
Design
Development
Others
Programming
Academic Credentials & Education
Formal degrees, graduate fellowships, and academic qualifications from premier institutions.
MSc, Computer Science and Engineering
Focus: Computer Science, Data Science, Machine Learning and Deep Learning
BA - Theology, Islamic Studies and History
Focus: Theology, Islamic Studies and History
Awards, Honors & Certifications
Academic honors, professional recognitions, and industry-validated certifications earned throughout the research and engineering career.
Interested in Research Collaboration or Systems Consulting?
Available for theoretical deep learning exchanges, high-throughput distributed systems advisory, guest seminars, or technical publication commissions.