I find, found & structure— find deals. found companies. structure everything in between. deals globally.
Cross-border by default. ↔
now (fall 2026): building WorkpaperIQ, investing through Taihill, and running a research program on the u.s. equity market's move to 23-hour trading.
tip: this page has tracked changes. hover the dotted lines.
Every generation gets one great repricing of the world. Mine got five.— so far. The tape is still running.
I became an investor at 18, in a world that still believed in one market. My first lesson was that nothing holds still.
Then the decade proved it. Tariffs went from 3% to a trade war that rerouted half a trillion dollars of supply chains. A pandemic cut the S&P 34% in 23 trading days — then delivered the fastest recovery in market history. The Fed took rates from 0 to 5.5% in eighteen months, repricing every asset on earth. Two economies that had spent thirty years converging began building parallel systems — separate exchanges, separate chips, separate clouds— and separate term sheets, which is where I come in. Then AI compressed a decade of change into every eighteen months.
I crossed the divide while it widened: ran a Financials book through the 2018 tariff swings and beat the benchmark by 1,500+ bps; graduated into the 2020 crash and kept buying; went deeper into digital assets in 2022, mid crypto-winter— when Bitcoin was down 75% and the seat was cheap. Timing is a personality trait., back when they were a leap of faith at a Big Four. I didn't cross it alone.[3]
What the decade taught me is simple: when the world fragments, the spread lives at the seams. Most people see borders as friction; I built my practice on the conviction that they're where value hides — if you can find the deal, and structure it to survive both sides. Which, in the end, is a judgment call[2] — and a relationship business.[1]
That's what I do. The rest of this page is the evidence.
The sixth repricing is the clock— the one institution I can't hedge, only study.. On December 6, 2026 the U.S. stock market starts trading twenty-three hours a day; clearing already moved to 24×5 in June. I've spent a decade inside the institutions that make markets work — audits, filings, financings, listings — and I now want to measure what happens when one of them changes while everyone watches. That's a computer-science problem as much as a finance one: how do you learn from an event stream whose rules change mid-stream? So I've started measuring it in earnest. The investing doesn't stop — it's where the questions come from.
Venture portfolio across fintech and deep tech — XtalPi (晶泰科技, HKEX: 2228), Lightelligence (曦智科技, HKEX: 1879), Databento, Revela. Fund II and III. This is where the questions come from.— and where the answers go back to work.
AI platform that reviews audit workpapers against PCAOB standards. In production with 7 IPO-stage pilot clients. Designed the product, the finding taxonomy and the evaluation pipeline; built the front-end with Claude. workpaperiq.com ↗
AI-native capital-markets platform preparing a Nasdaq listing. Advising on securities compliance and commercialization; supported the close of US$2M in pre-IPO financing.
Still the only position I have never diluted.
The U.S. equity market is becoming continuous, but its infrastructure changes discontinuously: NSCC clearing went 24×5 on June 29, 2026; the consolidated tape and Nasdaq's 21:00–04:00 session change on December 6; other venues follow on their own dates. That sequence lets us separate three mechanisms — clearing guarantees, quote transparency and venue competition — with stacked difference-in-differences and within-clock discontinuity designs, instead of one "after" dummy.
NightShift— the model runs while the market used to sleep. is the computational half: a regime-aware streaming model of order events with explicit session, trade-date and rule-state embeddings, judged on calibration, out-of-regime generalization and measured throughput rather than on trading profits.
interests: AI for finance and assurance · market microstructure under near-continuous trading · learning from asynchronous, non-stationary event streams · LLM analysis of corporate disclosures and audit evidence · verifiable AI for compliance workflows.
Manager, Financial Services Office (Boston; Toronto 2021–22). Digital asset research & AI teams; worked on the FTX bankruptcy and liquidation; from 2023 helped develop accounting-related AI applications on EY.ai; internal AI ambassador.
B.S. — Finance · Accounting · Entrepreneurship. GPA 3.7, Dean's List every semester. Teaching assistant, mathematics division. Babson College Fund: Financials/REITs sector +43.89% YTD vs benchmark.
Deloitte · Grant Thornton · Zhongtai Securities.
If your deal crosses a border, we should talk.
If your data has a clock in it, we should talk too.