“We ignore the public understanding of science at our own peril”
I come from chemical engineering (BUET) and work on neural networks, machine learning and optimization at HBit Labs. I’m also a full-stack developer, so most of my research ships as working software.
Latest: our paper Alkali-Mediated Hierarchical Defect Fingerprinting of Luminescent ZnO Nanohexagons for Multilevel Security Encoding is out in Materials Horizons (November 2026). I developed its deep-learning method, which identifies each banknote from a phone photo of its random luminescent spray pattern. Paper · Code
The common thread in my work is models that still recognise the right thing when real-world conditions change.
- Robust visual recognition: networks that hold up under new viewing angles, lighting and imaging modalities, from indoor drone navigation to medical imaging.
- Security and trust: anti-counterfeiting, network intrusion detection, and the safety and alignment of large language models.
- Machine learning and optimization for physical systems: AI for materials, wildfire risk mapping, and a non-linear drug-delivery controller for cancer treatment that combines SMC, ISMC and synergetic controllers tuned with a two-phase genetic algorithm.
- Alkali-Mediated Hierarchical Defect Fingerprinting of Luminescent ZnO Nanohexagons for Multilevel Security Encoding. Materials Horizons, 2026. Paper · Code
- Transferable Low-Rank Convolutional Bases for Onboarding Unseen Medical Imaging Modalities. arXiv, 2026. Paper · Code
- Reflex-Guard: A Low-Latency Guardrail for LLM Prompt Safety Using Dense Semantic Embeddings. arXiv, 2026. Paper
- BenSyc: Benchmarking Conversational Sycophancy and Human Alignment in LLMs for Bengali Contexts. arXiv, 2026. Paper
- Intrusion Detection Using Convolutional Neural Network: A Color Mapping Approach on NSL-KDD Dataset. NSysS 2024.
- A CNN-based Approach for Room Number Detection Using Drone in Indoor Environment. IEEE PEEIACON 2024. Code
- Development of a Remote Sensing and Machine Learning Based Spatial Fire Risk Assessment Framework for Wildland-Urban Interface: A Rohingya Refugee Camp Case Study. ICME 2023. Code
Citations and co-authors: Google Scholar · ResearchGate
- ResearchQ: a self-hosted, AI-assisted platform for systematic literature reviews.
- tessera: keyed tile scrambling for video that survives YouTube re-encoding, with a browser extension that descrambles it.
- Programming Languages: JavaScript, Typescript, Python, C, C++, Java
- Machine Learning & Neural Networks: TensorFlow, PyTorch, NeuralProphet, Darts, Transformers, Scikit-learn, Numpy, Pandas
- Web Development:
- Frontend: React, NextJS, Svelte, SvelteKit, HTML5, TailwindCSS, Bootstrap
- Backend: Django, FastAPI, NodeJS, ExpressJS, Spring Boot, Firebase, REST API
- Robotics: Robot Operating System - ROS, Arduino UNO, Raspberry Pi
- Software: ArcGIS Pro, AutoCAD, MATLAB, SPSS, Adobe Illustrator & Photoshop
Feel free to reach out via email or LinkedIn. I’m open to collaborations, code contributions, research discussions, and new project ideas.




