# Scandium Labs **AI Research for Computational Materials Discovery** Scandium Labs is an independent AI research organization focused on developing machine learning systems for computational materials science. Our work combines graph neural networks, scientific machine learning, and physics-informed artificial intelligence to accelerate the discovery and screening of advanced materials. Our mission is to reduce the computational cost of materials discovery by building models that complement first-principles simulations and enable scalable exploration of chemical and structural design spaces. --- ## Research Areas - Physics-Informed Machine Learning - Graph Neural Networks - Computational Materials Science - Crystal Structure Representation Learning - High-Throughput Materials Screening - Battery Materials Discovery - Scientific Machine Learning - AI for Scientific Discovery --- ## What We Build Our work includes open research, machine learning models, datasets, and tools for the materials science community. Current areas of development include: - Crystal property prediction models - Physics-constrained graph neural networks - Materials screening pipelines - Scientific datasets and benchmarks - Research software and reproducible training frameworks As our research progresses, models, datasets, and demonstration Spaces will be released through this organization. --- ## Featured Research ### PIGNet V2 **Physics-Informed Graph Neural Networks for High-Throughput Crystalline Material Property Prediction** PIGNet V2 explores the integration of physical constraints into graph neural networks for predicting electronic and thermodynamic properties directly from crystal structures. The project investigates multi-task learning for: - Formation Energy - Band Gap - Thermodynamic Stability The preprint serves as the initial proof of concept for the research direction that underpins Scandium Labs. --- ## Open Science We believe scientific progress benefits from transparency and reproducibility. Whenever possible, we aim to release: - Research papers - Model checkpoints - Training code - Datasets - Evaluation benchmarks - Technical documentation We welcome constructive feedback from researchers, students, and practitioners working in AI and materials science. --- ## Vision Scandium Labs is building AI systems that assist scientists in discovering the next generation of materials for energy storage, semiconductors, catalysis, and sustainable technologies. Our long-term objective is to develop reliable, physics-aware machine learning systems that become practical tools for scientific research and industrial materials discovery. --- ## Connect **Website** https://scandium-labs.com **GitHub** https://github.com/ScandiumLabs-in **LinkedIn** https://www.linkedin.com/company/scandium-labs/ **Email** team@scandium-labs.com --- ## Founder **Shamique Khan** Founder and AI Researcher Shamique's research focuses on graph neural networks, scientific machine learning, and physics-informed artificial intelligence for computational materials science. His work aims to bridge advances in modern AI with the scientific principles that govern materials discovery. --- *Scandium Labs is an independent research initiative dedicated to advancing artificial intelligence for scientific discovery.*