DataScribe: An AI-Native, Policy-Aligned Web Platform for Multi-Objective Materials Design and Discovery
Digital Discovery, 2026Department of Materials Science and Engineering, Texas A&M University
AI-native cloud platform unifying experimental data through knowledge graphs, FAIR-compliant metadata, and multi-objective Bayesian optimization for closed-loop materials discovery workflows.
— Unifies experimental and computational data into a machine-actionable knowledge graph
— FAIR metadata capture and schema/units harmonization
— Uncertainty-aware surrogate modeling with human-in-the-loop governance
— Closed-loop multi-objective Bayesian optimization for materials discovery
Materials informaticsFAIR dataOntologyKnowledge graphActive learningBayesian optimizationHigh-entropy alloys
View platform ↗TRAM-MP: Multi-Person Trajectory and Motion Reconstruction from Videos in the Wild
ArXiv Preprint, 2025Divyanshu Singh (Thesis Research)
Feed-forward method for reconstructing global trajectories and body motions of multiple people from monocular videos. Achieves comparable error reduction in trajectory estimation compared to leading models with speedup via transformer-based depth estimation.
— Speedup in camera estimation via feed-forward transformer
— Error reduction in root trajectory estimation vs. leading models
— State-of-the-art accuracy on multi-person benchmarks
— Generalizes to diverse human activities beyond MoCap-trained scenarios
Computer Vision3D ReconstructionHuman Motion CaptureDeep LearningMulti-Person TrackingSLAM
View repository ↗Efficient Parallelization of Point Cloud Registration: Spatial Indexing and Multi-Resolution
ArXiv Preprint, 2025Divyanshu Singh
Demonstrates significant speedup in classical ICP algorithms through CPU parallelization, multi-resolution refinement, and spatial indexing without specialized hardware.
— MPI-based parallelization with KD-tree spatial indexing
— Dynamic multi-resolution adjustment for registration accuracy
— Significant speedup over base ICP implementation
— No specialized hardware required
Point CloudICPParallel ComputingMPISpatial Indexing3D Processing
View repository ↗