Building AI that can reason about finance—and improve how research itself is done.
Selected AIDF and collaborative research spans live multi-agent financial forecasting, deep-document research, trustworthy retrieval, blockchain graph intelligence, human-behavior foundation models, multimodal scientific reasoning, synthetic stress testing and AI-native empirical methods. The common theme is simple: frontier AI should not only predict better; it should reason over complex systems, produce evidence, withstand scrutiny and make scientific conclusions more reliable.
× ScienceResearch Core
A portfolio organised around difficult research problems—not around one algorithm.
The map below shows the recurring research programs behind the publication portfolio. Select a theme to see what it is trying to solve.
Research
Agentic Financial Intelligence
Agents that search, reason, collaborate and execute multi-step financial research workflows.
Selected AI & computer-science conference research
These papers span frontier AI, data systems and computational research infrastructure—from LLM agents and financial benchmarks to blockchain graphs, behavior foundation models and multimodal scientific reasoning. Each summary is written for a broad audience, and every card links to Google Scholar.
AI-enhanced empirical methods, organizations and networks
The journal portfolio complements the conference work with durable research on AI-enabled empirical methods, decentralized-finance risk, networks and the consequences of AI for work: how to correct bias, build valid variables, control confounding and understand complex digital systems.
Research that travels between academia and real financial systems.
The academic portfolio is intentionally built around problems that matter in both worlds: autonomous research, evidence quality, model reliability, market forecasting, blockchain intelligence, human behavior and the statistical validity of AI-generated information.
From chatbots to research workflows
Our agentic work evaluates whether AI can search, synthesize and forecast—not simply answer isolated prompts.
Failure modes become measurable
Hallucination, retrieval drift and weak financial reasoning are turned into benchmarkable research problems.
Markets are graphs, not flat tables
Blockchain and exchange research models dynamic transaction structure, fraud, routing and arbitrage as connected systems.
AI should understand people—and improve inference
Behavior models, machine-generated measures, textual confounding and missing data all require stronger scientific safeguards.