Ian Cheung

Quantamental equities × applied AI

I build the tools between equity research and AI.

I work on both sides of the gap: close enough to equity L/S research to see where the process actually breaks, and technical enough in data integration and validation to build systems that fix it reliably.

Selected work Download CV
BSc Quantitative Finance, HKUST Hong Kong

01 — Selected work

Three tools, three failure modes.

Each project starts from something the research process does badly: missing the constraint, misreading the flows, or skimming the filings.

01

Bottleneck discovery

Polymer Capital Tech Expo 2026
Accepted submission

Bottleneck Screener

Find the constraint before the market names it.

A theme creates consensus exposure long before the constraint becomes obvious. The differentiated names usually sit deeper in the value chain — the suppliers whose lead times, backlogs, or capacity are quietly becoming the binding constraint.

The screener starts with a theme, scans conference material from that industry and from adjacent industries that touch it, maps the value chain, and quantifies each niche from IR filings. Bottlenecks are ranked; one is chosen for a deeper pass into the sub-chain and narrowed to specific names.

Wide enough to find the linkage. Narrow enough to act.

Pipeline
  1. 01 Theme input e.g. AI data centers
  2. 02 Cross-industry scan industry + adjacent conferences
  3. 03 Value-chain map from the niche outward
  4. 04 Filing quantification lead time, backlog, capacity
  5. 05 Bottleneck ranking constraints, not names
  6. 06 Deep dive sub-chain → specific names

OutcomeAccepted for Polymer Capital Tech Expo 2026.

02

Positioning & flows

Dragon Legend Capital
Quantamental Equities & AI

Taiwan Positioning Monitor

Read mechanical pressure, not just fundamentals.

Taiwan equities are unusually sensitive to mechanical flows: short balances, margin financing, insider activity, share pledges, and the split between local and foreign capital. Fundamentals can be right while price is set by something else entirely.

The monitor consolidates those signals into a single positioning view and flags divergences — institution against retail, foreign against local — where positioning is stretched and the tape is more likely to be driven by unwinds than by information.

Conceptual — flow cohorts
divergence foreign local aligned flow

OutcomeImplemented as a trading strategy at Dragon Legend Capital.

03

Fundamental research

Raffles Family Office
Equities & AI

Company Bottom-Up Reader

See what a skim of the filings misses.

A multi-agent pipeline over EDGAR that reads filings and transcripts end to end, then surfaces what changes before consensus notices: acceleration in a metric, divergence between metrics that should move together, and anomalies that demand an explanation. Findings are tied back to the events behind them.

Example SNDK: sales are growing, but NAND prices are decelerating. The margin slowdown is already in the data.

Built as research infrastructure, not a demo: deterministic calculations, schema-validated outputs, and every claim traceable to a primary filing.

Conceptual — signal attribution
acceleration divergence metric B metric A event

Outcome10 pitches; 4 accepted. Long MSFT and APH (+20%); trimmed APLD and SNDK (−20%).

Also built

  • HSI Rebalance Monitor Flags tracking-product liquidity against 30-day ADV around index changes.
  • AI Daily Brief Movers, drivers, and implications. Adopted for house views.
  • PE Portfolio Monitor Thesis catalysts and event alerts. Adopted for deal decisions.
  • Reporting Automation 30+ automated reports; 400+ cells and 4+ hours per month recovered.

02 — Experience

Research seats, engineering problems.

  1. Dec 2026 — Feb 2027

    Aspex Management Incoming

    Intern — Finance & Treasury Data, AI

  2. Aug — Oct 2026

    Dragon Legend Capital

    Intern — Quantamental Equities & AI

    Founded by ex-Macquarie PMs. Earnings-season strategy; daily index-flow and portfolio monitor; Taiwan positioning monitor, implemented as a strategy.

  3. Jun — Aug 2026

    Raffles Family Office

    Intern — Equities & AI

    Industry bottleneck skill, bottom-up anomaly reader, AI daily brief, and PE monitor; coached three department heads on no-code agents.

  4. Dec 2025 — Feb 2026

    Merit Asset Management

    Intern — Private Markets & Systems

    Automated fund and portfolio reporting; 400+ cells and four hours a month recovered; private credit and AI venture diligence.

  5. May — Sep 2025

    O-Oley Care

    Software Engineer

    iOS eye-tracking dementia screening; featured at HKUST Unicorn Day 2025.

03 — Approach

How I build.

01

Start at the bottleneck.

The useful tool begins with the step an analyst genuinely dreads or quietly skips. Workflow first, model second.

02

Primary sources, or it doesn't count.

Filings, transcripts, exchange data. Every output traces back to something that was actually said or filed.

03

Validate before it ships.

Deterministic math where possible, schema-validated outputs everywhere, and a fact-check on the claims. A tool that might be wrong is worse than no tool.

04 — Background

The rest of the picture.

Education

Hong Kong University of Science and Technology
BSc in Quantitative Finance, 2025–2029
CGPA 3.58 · Expected First Class Honours · Admission Scholarship

HKUGA College — DSE Best 5: 31/35. Founder of a HK$20K-funded app and the Finance Club.

Beyond the desk

Journaled equities since 2023 — paper account 18.8% p.a., live account 56.1% YTD (as of October 2026). Team lead for USD/KRW at Traders@UST. Trainer with Preface AI 5000.

Interests: neuroplasticity, indoor rowing, automobiles, predictive history — including an AI present-versus-past mirror.

05 — Contact

Let's talk.

I'm looking for seats where equity research and tooling sit together. If that's your desk, I'd like to hear about it.

Download CV