← Back to website For best results, choose A4 and disable browser headers and footers

SHI HONGYUE

Backend & Data Analytics Developer

Japan · Open to Remoteshihongyue2022@outlook.comgithub.com/shihongyue2022

Profile

Python-focused backend and data analytics developer with experience in financial time-series research, statistical modeling, and reproducible machine learning experiments.

Experience

Master's Research — Conditional Diffusion & Covariance Forecasting

Apr 2024 — Apr 2026

Tokyo University of Science · Japan

Designed Conditional Diffusion Score Matching (CDSM) and applied it to forecasting covariance structures in financial returns.

  • Conducted literature review, model design, Python implementation, experiment reproduction, and validation
  • Evaluated FX29, Industry49, and iShares14 against established covariance forecasting baselines
  • Demonstrated improved predictive accuracy and stability across empirical experiments

Teaching Assistant

2024 — 2026

University Course Support · Japan

Supported course delivery and student learning by adapting explanations to individual levels of understanding.

  • Answered student questions and provided supplementary explanations during class
  • Prepared model solutions, assisted with grading, and coordinated with instructors and other teaching assistants

Quantitative Finance Research

Sep 2015 — Jun 2019

Undergraduate Academic Project · China

Studied portfolio construction for index investing and implemented multiple approaches in Excel for empirical comparison.

  • Evaluated portfolio construction methods from risk and return perspectives
  • Built foundations in data structures, linear algebra, probability, statistics, economics, investment, C, and SQL

Projects

Conditional Denoising Score Matching (CDSM)

Python · PyTorch · Time Series · Statistical Modeling · Data Analysis

A Python research codebase for sliding-window covariance estimation using conditional diffusion in time and Fourier domains, evaluated against classical statistical baselines on financial time-series datasets.

  • Provides reproducible data preparation, training, baseline, metric, and visualization workflows for thesis and paper experiments

Equipment Inspection Management System (EIMS)

Python · Flask · SQLAlchemy · MySQL / TiDB · Docker · pytest

A role-based Flask web application where field engineers register equipment inspection results and administrators monitor equipment, assignments, and records in real time.

  • Built modular Blueprints, authentication and authorization, equipment/account/inspection CRUD, transactional services, MySQL and TiDB Cloud connectivity, Alembic migrations, Docker deployment, and HTTP integration tests