Quantitative finance · systems research

Singapore

Zheyuan Yu.

Market models.
Statistical tests.
Execution systems.

I build quantitative research tools and the systems that make their assumptions inspectable.

Explore the work Résumé

Current focusResearch, validation, and implementation

Up nextCambridge Computer Science, October 2026

Built to
be tested.

Selected work in market mechanics, machine learning, and decision-making.

Execution systemsC++20 / WebAssembly · 2026

Orderbook

A market is a queue.
Build the engine that keeps it moving.

A C++ matching engine tested with one NASDAQ TotalView-ITCH session.

Historical validation
286,725 / 286,725

Included executions agreed with the visible front of the queue across 10 maintained books.

Parser study
268.7M messages / 21.2 s

One TotalView-ITCH file. This is parsing throughput, not engine latency.

Read case study Open live build Source code

Loads the WebAssembly build into this frame.

preview · static ladder, illustrative sizesOpen full demo ↗

2019-12-30 session. Cross-path executions and opening-process releases excluded. Browser demo uses a simulated market, not historical replay.

Market GNN / 2026

Connected
doesn’t mean
predictive.

Do relationships between stocks add return-predictive signal? Twelve leak-controlled experiments found little evidence.

Read case study Source code
13F co-holding lead-lag90 stocks · 2014-2024

Not statistically
significant.

HAC t-statistic 1.4

Evaluation Purged walk-forward

The detection threshold is not a confidence interval or evidence of zero effect. Pinned results ↗

Interactive learning / 2026

A better
next hand.

Poker Trainer

Learn a concept. Drill the spot. Play the hand. Then find out what to work on next.

Open live build

Grades distinguish solver-backed references, hand-built charts, and heuristic estimates.

A poker training spot moves through learning, practice, and feedbackAn illustrative decision path begins with a spot, branches into the Learn, Trainer, and Play modes, then returns an evidence-labelled grade. No solver output or hand recommendation is shown.POKER TRAINER · DECISION SURFACEOne spot, three ways to learn from it

Figure 03 · 20 interactive lessons · three table formats · server-authoritative private rooms

The app states whether a grade is Nash-anchored, based on a hand-built chart, or a heuristic estimate.

Open the trainer

Private research prototype / 2026

Quantnews

From the day’s filings
to the next research question.

Read case study

RSS / SEC filings / Insider trades
Earnings calls / Analyst revisions

  1. Classify events
  2. Group historical outcomes
  3. Filter a research list

Workflow only. Targets, splits, costs, baselines, and results are private.

Earlier exploration 2022

Longest Simple Path

Graph learning for an NP-complete problem. Unpublished high-school research; generalization is not validated.

View archive

Experience

Ideas into
working
systems.

Research, engineering, and the distance between the two.

The longer story
  1. Apr 2026 - PresentCurrent role

    MS Capital

    Quantitative Research Intern

    Implemented factor-momentum ideas from the literature and automated the path from market data to out-of-sample backtests.

  2. Jan 2026 - Mar 2026

    Huawei Singapore

    Computer Vision Algorithm Engineer Intern

    Built region-of-interest extraction and explored a hybrid VAE and reinforcement-learning approach to image compression.

  3. Oct 2023 - Dec 2023

    Temasek International

    Data Analytics Intern

    SQL and Tableau analysis for the investment team; dashboards the senior analysts adopted.

  4. Nov 2022 - Dec 2022

    NUS December Course

    Lecturer

    Designed curriculum and original contest problems for 100+ secondary-school competitive programmers.

Competitive record

Algorithms, mathematics, and a lot of practice.

Grandmaster.

Codeforces Grandmaster

Peak rating 2488

Competitive Programming

  • USACO Platinum 1000 / 1000, Dec 2022

  • Google Code Jam

    Top 50 worldwide, 2022

  • Meta Hacker Cup

    Top 50 worldwide, 2022

  • National Olympiad in Informatics, Singapore

    Gold 2021 & 2022, Third Runner-Up 2023

  • Asia-Pacific Informatics Olympiad

    Silver 2023, Bronze 2022

Mathematics

  • USA Mathematical Olympiad

    Qualifier, 2021

  • USA Junior Mathematical Olympiad

    Qualifier, 2020

  • Singapore Mathematical Olympiad, Open

    Top 10, 2020

  • Singapore Mathematical Olympiad

    Gold, 2020-2022

Scholarships

  • Tay Eng Soon Scholarship

    NUS High School

  • Edusave Scholarship for Independent Schools

Full record in résumé
Zheyuan Yu speaking into a microphone
Zheyuan Yu Singapore

Behind the work

Algorithms first.
Markets next.

I came to market research through competitive programming, then worked on data and machine-learning systems at Temasek and Huawei. I now do factor research and backtesting at MS Capital, before starting Computer Science at Cambridge in October 2026.

Starts Oct 2026

University of Cambridge

BA Computer Science, St John's College · expected 2029

2020 - 2023

NUS High School of Math and Science

Honors in Mathematics; Majors in CS, Physics, Chemistry