Hello! 👋

I'm Henry (Haorui) Zhang

Data Scientist

About Me

I am someone who works at the intersection of mathematics, computer science, data science, and quantitative finance. I am currently a Data Science Intern at Tencent (Level Infinite, its global gaming brand) in Palo Alto, where I apply statistical modeling, machine learning, and A/B testing to turn large, complex datasets into data-driven insights across game development, marketing, and operations. Alongside this, I am pursuing graduate studies at the University of Pennsylvania and UC Berkeley, and I am passionate about data modeling, machine learning, and AI-driven systems. I enjoy building practical, scalable solutions and exploring how intelligent models can solve real-world problems.

Programming Languages

Python C/C++ Java

ML/DL Frameworks

Scikit-Learn Pandas TensorFlow PyTorch XGBoost CUDA

AI Skills

RAG LlamaIndex LangChain Ollama vLLM Prompt Engineering LLM Inference Vector Retrieval Agent Workflows

Tools

R MATLAB STATA VBA Excel Git Docker Power BI Tableau LaTeX

Databases

MySQL PostgreSQL MongoDB Neo4j

Big Data & Cloud

Apache Spark Databricks Snowflake AWS GCP Azure

Education

University of California, Berkeley

Master of Analytics

Aug 2025 - Aug 2026Berkeley, CA
  • CGPA: 3.968/4.0
  • Coursework: Optimization, Data Analysis, Machine Learning, Database, Financial Engineering System

University of Pennsylvania

Dual Degree: M.S.E. in Data Science & Computer and Information Technology

Aug 2023 - May 2026Philadelphia, PA
  • CGPA: 3.94/4.0
  • Coursework: GPU Computing for ML Systems, Artificial Intelligence, Natural Language Processing, Deep Learning

Washington University in St. Louis

M.S. in Quantitative Finance

Aug 2023 - Dec 2024St. Louis, MO
  • CGPA: 4.0/4.0 (Ranking: 1/89)
  • Honors: Outstanding Finance Student, Charles F. Knight Scholarship, Beta Gamma Sigma Honor Society
  • Coursework: Mathematical Finance, Real Analysis, Stochastic Analysis

Sun Yat-sen University

B.Econ. in Finance

Sep 2019 - Jun 2023Guangzhou, China
  • CGPA: 3.9/4.0

Experience

Data Science Intern

Tencent

Jun 2026 - PresentPalo Alto, CA

    Artificial Intelligence Research Intern

    Global Key Advisors

    Oct 2025 - Jan 2026San Francisco, CA
    • Built an automated pipeline for noisy, multi-format financial records across 1,000+ startups, converting unstructured text into structured corporate intelligence
    • Derived 20+ risk factors and event signals from financial disclosures through LLM-driven context analysis
    • Backtested LLM-derived signals and achieved a 25% higher return than traditional machine learning baselines in explaining cross-sectional variation

    Deal Sourcing and Client Coverage Intern (Quantitative Analysis Role)

    Haitong Securities

    Jul 2022 - Oct 2022Qingdao, China
    • Conducted fundamental analysis on financial statements of 200+ corporations with cumulative assets exceeding RMB 500 billion, successfully flagging a high-risk issuer before potential credit events
    • Developed Python and MathPix automation to digitize financial tables in 15 seconds per table, saving around 40 team-hours weekly by removing manual entry
    • Analyzed debt structures across 300+ companies and supported data-driven screening for investment banking debt issuance targeting

    Quantitative Researcher Intern

    Tsingtao Stone Asset Management

    Dec 2021 - Mar 2022Qingdao, China
    • Conducted strategy development/back-test of energy futures trading, eliminated 30% of underperforming models
    • Analyzed 12 financial factors, highlighted consistent significance of SMB and HML factors
    • Performed quantitative analysis for 20 indices, identified root causes of model-market discrepancies

    Projects

    02

    Distributed GPU Computing for Machine Learning Systems

    Engineered a high-performance PyTorch-style tensor and autograd engine in C++/CUDA, achieving 2.3× faster convolution and 3.7× faster pooling than LibTorch on megapixel-scale inputs. Built a full GPU training stack with custom-optimized kernels for convolution, RNN/LSTM, normalization, pooling, and self-attention, enabling 1–7 ms forward/backward passes. Designed distributed LSTM/ML forecasting pipelines using NCCL-based data parallelism, delivering a 25% error reduction over baseline models in financial backtests.

    C++CUDAPyTorchcuBLAScuDNNAWSNCCL
    03

    LLM-Driven Financial Derivative Analysis System

    Engineered a financial reasoning system integrating GPT-4 and Claude Sonnet APIs to analyze complex derivative pricing scenarios, applied chain-of-thought, few-shot, and instruction-tuning strategies for a 55% performance lift over zero-shot baselines, and outperformed traditional pricing models by 40% and ML baselines by 15-60%.

    PythonLLMGPT-4Claude SonnetPrompt Engineering