Built in the open, with 40 years of practitioner depth at the intersection of quantitative finance and emerging technology.
I wrote my first neural network in 1988 — on a 286, in a commodity pool office, trying to find signal in futures data. Since then I've shipped production systems across every wave: expert systems, MPT-based risk engines at Ramcap, managed-futures models at Navigator Fund (twice noted in the Wall Street Journal), curriculum at the College for Financial Planning, and now fuzzy-logic + ML hybrid systems for crypto grid trading. Same thread through all of it: quantitative finance meets whatever compute just became possible.
2026 is not a pivot. It's the year the compute finally caught up to ideas I've been tracking since the Reagan administration.
Mamdani and Sugeno architectures for decision support where the output has to be explainable to a compliance officer, not just accurate.
Fuzzy rules encode the expert's intuition; machine learning tunes the membership functions from data. Neither pure black-box nor pure hand-coded.
Regime detection, signal extraction, and portfolio-scale backtesting.
For regulated contexts — fiduciary, EU AI Act, SEC-examinable. The output includes a defensible English rationale, not just a number.
Software that sits next to a human, not software that replaces one. Fuzzy outputs in plain English, sized to the firm's risk budget.
Reproducible, peer-reviewable, with the code and the data paths documented.
A crypto grid-trading analysis app running fuzzy-logic risk throttling in production. Not a research demo. Live users, live capital, live compute. The fuzzy engine tunes position sizing based on volatility and drawdown regime — and every decision logs an English-language rationale.
The umbrella research house publishes methodology papers on fuzzy logic, post-MPT portfolio construction, and hybrid ML architectures. Paper 2 (in progress) documents the fuzzy membership-function design and the reproducibility artifacts behind the CoinRoc production engine.
A decision-support workflow for registered investment advisors who want a crypto sleeve they can actually explain to a client and a compliance officer. MVP ships in 5–7 weeks. Fuzzy inference outputs map directly to the DOL safe-harbor 6-factor framework — so the advisor's recommendation memo writes itself.
Who want a crypto or alternatives sleeve their compliance officer can approve, not a black box their clients can't understand.
Looking for quantitative infrastructure that's built by a practitioner who's run money, not by a bootcamp grad who's read about it.
Who need applied AI capacity — fuzzy inference, hybrid models, explainability layers — without hiring a ten-person research team.
Evaluating AI strategy who are tired of deck-driven consultants and want someone who ships code.
If you're building in one of those lanes and you want a conversation with a practitioner — not a pitch, not a deck, a conversation — reach out. I take on a small number of advisory engagements each year, and I'm selective about fit. The work has to be interesting and the problem has to be real.
