Skip to content
Fengrui Yang

Fengrui YangUniversity of Toronto

I build systems that turn messy data into decisions.

Industrial Engineering, Machine Learning, Quantitative Research

Selected WorkResume (opens in a new tab)
LinkedIn (opens in a new tab)
System loopfig. 01
  1. 01Datamessy, partial, drifting
  2. 02Modelestimate, learn, optimize(emphasized for the selected role)
  3. 03Decisionan action under constraints
  4. 04Validationdoes it survive reality?(emphasized for the selected role)

Validation feeds back into the data and the model.

A model is only useful if the decision survives reality.

Evidence

Quant systems (supporting for this role)
Research, then portfolio, then executable decisions
ML validation (primary for this role)
Baseline-first, out-of-sample evaluation
Physical systems
Experimentation, modeling & optimization
Production software
Software built under real delivery constraints

Selected Systems

03 systems · ordered for Machine Learning

  1. 02. Quantitative ML

    Quantitative ML Research

    Testing whether machine-learning models produce useful predictive signal beyond simple market baselines.

    Relevance · Machine Learning

    Emphasizes target construction, baseline selection, model comparison, out-of-sample evaluation, and failure analysis.

    Problem
    A model can produce predictions without producing meaningful predictive value.
    System
    1. Target definition
    2. Baseline
    3. Model
    4. Out-of-sample test
    5. Failure analysis
    Evidence
    • Baseline-first evaluation (emphasized for this role)
    • Explicit target construction (emphasized for this role)
    • Out-of-sample testing (emphasized for this role)
    • Comparison of linear and nonlinear approaches
    • Emphasis on model failure as useful evidence (emphasized for this role)
    Role
    Research, Modeling, Validation

  2. 01. Personal system

    View case study

    Quant Research & Execution Platform

    An end-to-end quantitative research system connecting research evidence, portfolio decisions, market evidence, and execution.

    Relevance · Machine Learning

    Shows how model or research outputs must survive data lineage, decision construction, and downstream execution constraints.

    Problem
    How can research, portfolio construction, market evidence, and execution remain consistent as one decision process?
    System
    1. Research
    2. Portfolio
    3. Reference price
    4. Execution
    Evidence
    • Reproducible research pipeline (emphasized for this role)
    • Point-in-time evidence (emphasized for this role)
    • Portfolio decision workflow
    • Execution-aware system design
    Role
    Architecture, Research, Implementation
    Stack
    Python, PostgreSQL

  3. 03. Engineering system, Team engineering project

    Dynamic Adaptive Mooring System

    An interdisciplinary engineering project studying how an adaptive mooring system can respond to changing wave conditions.

    Relevance · Machine Learning

    Provides a physical-system setting where modeling must remain grounded in measured system behavior and experimental validation.

    Problem
    How should system response be measured, modeled, and optimized under changing physical conditions?
    System
    1. Wave conditions
    2. Sensors / experiments
    3. System response
    4. Analysis
    5. Optimization / validation
    Evidence
    • Physical experimentation (emphasized for this role)
    • Sensor and DAQ data
    • System modeling (emphasized for this role)
    • Performance analysis
    • Validation under changing conditions (emphasized for this role)
    Role
    Experimental data analysis, Modeling, Optimization

Experience

2025 — 2026

  1. Quant Finance & ML Intern

    Global AI

    Conducted financial ML research across historical market-data workflows, baseline-driven out-of-sample evaluation, and quantitative strategy research.

    • Research
    • Data systems
    • Model validation

  2. Front-End Developer

    Wuhan Tiancai Shentong, PEY work term

    Delivered and maintained front-end functionality during a year-long PEY work term, with implementation, testing, and iterative software delivery.

    • Front-end engineering
    • Software delivery
    • Testing

About

My work tends to sit where models meet real decisions.

I care not only about whether a model performs, but also about the data feeding it, the constraints around it, how it is validated, and what happens downstream.

Industrial Engineering gives me the systems and optimization perspective; machine learning and quantitative research give me tools to model, test, and improve those systems.

Education

University of Toronto

Bachelor of Applied Science, Industrial Engineering

Minor in AI in Engineering

Expected 2027

Technical Capabilities

  • Modeling & Validation (primary for this role)

    Capabilities
    Target design, Baseline construction, Out-of-sample validation, Experiment design
    Tools
    Python, scikit-learn, LightGBM, statsmodels, TensorFlow / Keras

  • Quant & Optimization

    Capabilities
    Factor research, Portfolio construction, Time-series analysis, Optimization, Operations research
    Tools
    cvxpy

  • Data Systems

    Tools
    NumPy, pandas, PostgreSQL, SQLite, MongoDB, Spark, FastAPI

  • Software Engineering

    Tools
    TypeScript, JavaScript, React, Next.js, Java, Spring Boot, Docker, Git / CI

Contact

Interested in opportunities across machine learning, quantitative research, and data / analytics.