M.Sc. in Computational Sciences · Laurentian University, Canada · CGPA 9.10/10

I build end-to-end ML systems — from watershed cell segmentation pipelines to RAG-based retrieval architectures — and ship them as production deployments. My work sits at the intersection of applied deep learning, NLP, and backend engineering.

Currently pursuing research in adversarially robust RAG pipelines for intrusion detection (targeting IEEE TDSC) while preparing MalariAI for journal submission to Elsevier CMIG.

Accurate Prediction of Pulmonary Fibrosis Progression Using EfficientNet and Quantile Regression Rofiqul Alam Shehab, Kaysarul Anas Apurba, Md. Ahsanuzzaman, Tanzilur Rahman IEEE TENSYMP 2023 · Canberra, Australia

A two-stage decoupled framework solving three persistent failure modes in malaria blood smear analysis.

Python PyTorch OpenCV EfficientNet HuggingFace

RAG-based pipeline for natural-language querying over scientific literature — chunking, embedding retrieval, and LLM-grounded answer generation with source attribution.

Python LangChain Vector DB HuggingFace RAG

Multi-agent RAG architecture for network intrusion detection robust against retrieval poisoning and prompt-injection attacks — trust-score filtering, semantic anomaly detection, and prompt sanitization. Targets the security robustness gap identified across existing RAG-IDS taxonomies.

Python RAG LLM Security Multi-Agent CICIDS2017

Full-stack job application tracker — React (Vite) + Tailwind on Vercel · Django REST + Gunicorn + Nginx on AWS EC2 · JWT auth · Google OAuth · React Native mobile app

Interactive AES vs RSA visualiser — live encrypt/decrypt demos, performance benchmarks, and security analysis. Built with React and Tailwind CSS.

ML / AI

Backend & Web

Cloud & DevOps