I earned my Ph.D. in Operations Management at NYU Stern, advised by Jiawei Zhang and Divya Singhvi. My work sits at the intersection of online learning, stochastic optimization, and LLM-agent systems for quantitative research. On the theoretical side, I develop adaptive online gradient descent methods for non-stationary environments, together with primal–dual algorithms for large-scale resource allocation; my thesis research is motivated by and applied to problems in inventory control, e-commerce order fulfillment, portfolio selection, and related online convex optimization.

In parallel, I run AlphaBot, a multi-LLM agent system for systematic alpha discovery. The architecture treats frontier LLMs as orthogonal proposers and a backtesting harness — calibrated to hedge-fund-grade quant-research review — as the bar that decides what survives. The language models propose; the harness disposes. AlphaBot has been validated on two markets: US equities at Cubist (Point72) during my Sep 2025 – Feb 2026 research engagement, where it identified 80+ significant meta-alphas across momentum, mean reversion, liquidity, and information flow; and an independent self-funded crypto deployment outside Cubist with public live performance over 18 months. One research finding from running the same task across three frontier model families: research-output quality ranks Opus > GPT > Gemini.

IvorySquare (Apr 2026 – present, with Han Yan) treats peer-reviewed methodology — across finance, accounting, economics, and operations research — as a first-class tool surface for LLM agents. Skills are paper-derived, citation-grounded, and gated by purpose-built evaluation harnesses; the human-expert layer remains disjoint from engineering through declarative persona contracts. The skill graph is structured in two coupled tiers — a foundational concept layer at textbook-subsection granularity under prerequisite ordering, and a paper-derived methodology layer with formal implementations and line-item citation provenance — both exposed as declarative MCP and OpenAI tool surfaces.

A research direction I find particularly promising is the use of academic citation networks as a structured post-training substrate for tool-using LLMs. The topology of how ideas build from primitive methods to composite ones is a natural curriculum; a paper-indexed library of verifiable, test-backed skills turns that curriculum into supervised tool-use data.

Before NYU, I received a B.Sc. in Mathematics from the Chinese University of Hong Kong with First Class Honours. I was also a Quantitative Research Summer Intern at Optiver US, working on 0-dte SPXW volatility-change-rate estimation.

I am based in New York City. Email: sy2576 [at] stern [dot] nyu [dot] edu.

Research Interests

LLM-Agent Systems and Tool-Using AI

• Multi-LLM agent pipelines and closed-loop agent evaluation
• Tool-using LLMs and post-training for domain-specialized capability
• Evaluation, hallucination mitigation, and trustworthy AI

Quantitative Finance and Alpha Research

• Systematic alpha research and explainable factor discovery
• Multi-market validation of agent architectures (equities, crypto)

Online Learning, Stochastic Optimization, and Sequential Decision-Making

• Adaptive online stochastic gradient descent and regret minimization in non-stationary environments
• Primal–dual algorithms for online resource allocation (inventory control, e-commerce fulfillment, revenue management)