Hongjin He

Education

B.Sc., Risk Management & Business Intelligence — HKUST, Extended Major in Artificial Intelligence · Honors Track in Mathematics

Sep 2025 – 2029 (exp.)

Core GPA 4.15 / Cumulative GPA 3.7
Coursework: Honors Calculus & Probability, Discrete Mathematics, Stochastic Processes, Convex Optimization, Microeconomics

Visiting Student, Computer Science — Stanford University

Summer 2026

STATS 217 (Stochastic Processes), CS 229 (Machine Learning), EE 364 (Convex Optimization)

Experience

Co-Founder & Head of Quantitative Strategy — Alpha Flow

Sep 2025 – Present

Lead theoretical design of a world-model framework for quantitative finance: an SDE-based latent state formulation (encoder–world-model–controller) extending Ha & Schmidhuber-style world models to market dynamics, with mean-field-game and stochastic control components
Authored a mathematical framework covering Itô-integral state evolution, hierarchical mean-field games, and Lyapunov stability analysis of regime transitions; drafting a theory paper targeting quantitative-finance journals
Direct a small cross-functional team (AI research, quant strategy, risk) building toward an investable systematic strategy

Quantitative Research Consultant — WorldQuant (BRAIN platform)

Mar 2026 – Present

Designed and submitted systematic alpha signals using diffusion-based generative search over factor structures, achieving IC = 0.148 on out-of-sample evaluation
Built a modular, real-time backtesting framework in Python to validate signal stability and turnover before submission

Research Assistant, UROP — HKUST (PI: Prof. Tony Cho)

May 2026 – Present

Study the robustness of LLM-extracted alternative-data signals from unstructured financial disclosures (SEC 10-K filings, earnings releases) as inputs to quantitative models, scaling from a 30-firm pilot to 3,258 U.S. public firms and 180,000+ model responses; target venue: Journal of Accounting Research
Re-implemented SelfCheckGPT (RoBERTa BERTScore, LLM-as-judge, n-gram perplexity) from scratch after auditing the inherited codebase and finding a critical measurement bug — a length-based heuristic mislabeled as semantic similarity
Designed a checkpointed, GPU-batched pipeline on an HPC/Slurm cluster with async concurrent API calls; empirically characterized nonlinear rate-limit behavior across API keys and tuned concurrency accordingly
Applied ANOVA and variance decomposition (η²): question cognitive demand explains 31× more inconsistency than firm size, evidence that signal reliability is task-structural rather than purely data-quality-driven

Co-Founder, Quantitative Finance Arena — GBA Redbird Entrepreneurs Alliance

Nov 2025 – Present

Co-built a quantitative trading strategy platform with Cantonese voice-input support, the first of its kind in the initiative

Selected Projects

MicroWorld — Causal, Multi-Agent World Model for Markets (650+ GitHub stars)

Core project of Alpha Flow: models market prices as the emergent equilibrium of interacting, utility-maximizing investor-agents (causal structure) rather than fitting a single stochastic process to correlations — targeting factor decay at its root
Framework parallels the derivation paradigm of the 2026 Fields Medal (Yu Deng, with Hong Wang) — N-body Newtonian mechanics → Boltzmann equation — mapped to N investors under utility maximization → a Fokker–Planck–Kolmogorov equation over market state via a McKean–Vlasov SDE limit

multi-agent world modelscausal inferencemean-field gamesMcKean–Vlasov SDEs

Mathematical Framework for World Models in Quantitative Finance

SDEmean-field gamesstochastic controlLaTeX

Diffusion-Based Alpha Mining

Generative search over discrete factor structures for robust, low-correlation alpha discovery on noisy financial time series

Real-Time Quantitative Backtesting Framework

Modular Python framework for evaluating systematic strategies against live and historical market data

Honors & Leadership

Top 92 nationally, Tsinghua University Physics Climbing Program (Gold Medal–equivalent, Chinese Physics Olympiad track)

2024

Rated "Excellent" (top tier), Tsinghua University Physics Bridging Program — earned direct-admission recommendation

Jul 2024

Host & Panel Guest, Future Science Prize AI Roundtable Forum — joint discussion with Prof. Mengdi Wang (Princeton, Director of AI Innovation Center), Prof. Wei-Ying Ma (President, Hong Kong Institute of AI and Science; ACM/IEEE Fellow), and Prof. Tim Kwang-Ting Cheng (Vice-President, HKUST)

Nov 2025

Hongjin He 何泓锦

Education

Experience

Selected Projects

Honors & Leadership

Skills