Md. Saidul Islam
M.Sc in Materials Science & Engineering (CAU, Kiel), focusing on data-driven methodologies & materials analytics. Able to build or reproduce (from existing research) ML pipelines, FEM models, and research apps that move from notebooks to usable tools. Proven & recognized Rapid familiarization skill to a new topic, enabling effective cross-disciplinary collaboration and innovation. Here, I demonstrate some of my works at the interface of materials science and AI.
Selected Works
Themes
- Melting-point prediction: 2-level custom-stacking (RF/XGB/LGBM/MLP) achieved R² ≈ 0.83 on ~3.041k records.
- Oxidation-state assignment: soft-voting ensemble on ~7k OQMD samples reached R² ≈ 0.91 (proxy metric), delivering competitive quality on ~15× less data than typical literature sets.
- Semantic band-gap knowledge graph: RDF/SPARQL schema parsed (primary task)/queried (secondary task) via a local LLM (llama3.2:3b) for explainable lookups.
- Developed a proof-of-concept materials database app with auto ETL and Fly.io deploy, enabling end-to-end tracking from raw data to model results.
Toolbox
Programming & Data
PythonMATLAB
SQLHTML
CSSRegEx
C++Fortran
AI / ML
EnsemblesPINN
SINDyAutoMLDNNs|GNNs
SHAPGenerative AI
RDF/SPARQLActive Learning
Materials Informatics
MatminerPymatgen
RDKitOQMD
Materials ProjectSemantic WebSchNet
Simulation & FEM
COMSOLAbaqusASE
SimScaleLAMMPSOVITO
Characterization
AFMTEM
SEMXRD
VSMDSC/TGA
Web & Databases
FlaskJinja
DockerSQLite
Focus areas
- High-Throughput Materials Development workflows
- Smart / Functional Materials
- Heat/Corrosion-resistant high-strength alloy systems
- AI in Materials Science (interpretable models & GenAI)
- FEM for design & validation