Recent work is intentionally summarized at a high level. Earlier projects include more context and links to the publications, tools, and public resources behind them.
Building agentic AI pipelines for scientific and engineering discovery by integrating structured evidence, networks, embeddings, and expert judgment. A complementary line of research uses embedding models to map technology trajectories and predict capabilities across materials and device platforms.
Building large-scale, validated links between funding acknowledgments, organizations, and publications. The broader research program examines how public, philanthropic, and corporate support shapes scientific directions, researcher trajectories, and the geography of knowledge production.
Developing embedding models that represent design choices and technical attributes extracted from papers and patents. The research examines whether interpretable directions in these spaces align with measurable progress, expose important tradeoffs, and help predict how technological capabilities evolve across materials and device platforms.
A browser-first platform for interactive graph visualization and embedding exploration. The current release combines a WebGPU renderer with WebGL2 fallback and a WebAssembly graph store, supporting real-time layouts and analysis at million-node scale across browsers, notebooks, desktop workflows, and agentic tools.
Developing network and embedding methods to study how researchers and fields change direction, recombine ideas, and approach new frontiers. The work relates paper trajectories to collaboration structure, scientific impact, disruption, and the conditions that help new directions emerge.
Studying coordinated behavior, information integrity, and public discourse across emerging social platforms. Current work combines network structure, text embeddings, topic models, and human validation to understand how collective behavior forms and changes online.
Developed a network-based method for detecting statistically significant long-range relationships between temperature and precipitation anomalies. The project used Granger causality to reveal climate teleconnections while explicitly controlling for indirect effects.
Mapped the changing scale, collaboration structure, funding, and topical reach of landmark physics, astronomy, and biomedical research programs. The public project includes reproducible datasets, processing code, and interactive institution and science maps.
Applied network science, machine learning, and scientific-data pipelines to problems in single-cell genomics, wound-healing materials, neuroscience, and biological networks. The work ranges from robust cell-type identification and literature-scale synthesis to reproducible cloud workflows, showing how shared computational representations can support very different biomedical questions.
Contributed network-analysis and real-time visualization infrastructure for studying online information diffusion and coordinated activity. This work connected research pipelines to accessible browser tools used by scientists, journalists, and public-interest investigators.
Developed multiscale measurements for describing nodes and networks through concentric structure, accessibility, symmetry, and local dimension. The methods were applied to collaboration, text, biological, urban, transportation, and infrastructure networks and released with supporting software.
Studied how recent attention predicts citation bursts, how collaboration patterns relate to productivity and visibility, and how stable scientific communities remain as citation networks evolve. The work introduced both empirical findings and new measures of community robustness.