Md. Saidul Islam

I am a materials scientist (M.Sc., CAU Kiel) working at the intersection of materials science and AI. My master's thesis investigated nonlinear dynamics in implantable magnetoelectric sensors: SINDy discovered the governing equations from acquired signals (R² ≈ 0.99), with ANN models serving as accuracy benchmarks — an early commitment to interpretable, physics-grounded modeling over black-box prediction.

From there I moved into materials informatics: ensemble models for property prediction, a PINN for elasticity field inference, and a semantic (RDF + SPARQL) knowledge graph with LLM-assisted batch/single-parsing with provenance from different sources (web, pdf, natural language) & querying — each project targeting a specific gap between raw materials data and actionable insight.

The current focus is atomistic simulation as a first-principles data source: parametric multivariable MD study of silicon nanowire's mechanical response (LAMMPS) and First-Principle DFT study of strain-driven electronic transitions (the strain dependendent Dirac Cone) in graphene (Quantum ESPRESSO), feeding toward ML interatomic potential development. The longer arc is a DFT → MD → ML pipeline for scalable, physics-informed materials modeling.