Scientific AI & Quantitative Modeling
Investigating non-Euclidean graph topologies, multi-LLM consensus architectures, and quantitative risk mechanics. Grounding theoretical AI in verifiable empirical systems.
Developing GraphSAGE and Graph Attention Network (GAT) algorithms to analyze directed non-Euclidean ecological trophic networks, complemented by multi-LLM data synthesis pipelines across Gemini, DeepSeek, Anthropic, and OpenAI to unblock literature extraction.
Investigating how relational inductive biases, message-passing formulations, and node-level attention enable machine learning models to reason over complex interconnected graphs.
Architecting multi-agent ensembles that query disparate frontier foundation models, reconcile conflicting outputs through semantic voting, and produce verifiable structured representations.
Applying continuous-time finance, stochastic differential equations, and asset-liability matching (NSFR) to real-world credit origination and automated financial systems at WorldQuant University.
Associated Technical Notes & Reflections