ABOUT

One engine, many directions — and a public trail of what failed.

Blaque Baux is an open, Julia-based systematic-research initiative: a single governed engine steered into a family of strategy sleeves, each pointed at a different corner of the market.

Every idea is tested against the same honest bar. Most do not clear it — and instead of hiding them, we keep them public as a research graveyard: the failed thesis, the constraint, or the risk finding stays visible so it can inform the next test. The handful that survive are built into governed drivers and validated before they are ever described as live. The corpus is the whole record, kept open on GitHub — live systems, work in validation, active research, and null results side by side.

The standard: publishing the failures is the point.

Anyone can point an AI at a market now. The edge is understanding what the data actually says — and being willing to show your losses. That discipline runs through everything here:

• The red scorecard is the pointA rejected strategy, documented with its reason, is a result — not an embarrassment. Most of the corpus is exactly this.
• Validate before liveNothing is called real on a backtest. It earns a governed paper track, then real money — or it stays research.
• The tail is the whole storyReturns are fat-tailed, so a bare Sharpe misleads. We test normality first, and weight the crash as heavily as the average.
• Flag the gap, don't fake itWhere the data isn't there, we say so. No proxy dressed up as the real thing.

Governance is the product.

The platform is built in Julia around a venue-agnostic execution controller and a Layer-3 live-money safety gate — preflight checks, idempotency, reconciliation, a high-water-mark and a kill switch. The research may use modern tooling, but no language model sits in the order path: every allocator emits a target book, and reproducible code routes it through the gate.

Performance is judged with a fat-tail toolkit, not a headline number: Jarque–Bera for normality before any mean-variance claim, Jensen's alpha and M² instead of Sharpe alone, and a crisis-correlation check — how a stream behaves on the market's worst days, the 2008 tell — before anything is trusted to diversify. The point is not a clever signal; it is a book you can actually run.

Carter Warrens.

Blaque Baux is a subsidiary of Carter Warrens — a boutique firm built on a simple observation: the execution capabilities that power the world's largest institutions shouldn't be reserved only for them. Carter Warrens brings that institutional-grade discipline — program management, transformation, and high-stakes delivery — to growth-stage companies, private funds, and public-sector organizations.

Its partners carry 40+ years of combined program leadership at global banks and market-infrastructure firms, based in the New York metro. Blaque Baux is where that discipline is turned on the markets themselves: systematic research held to an institutional bar, in the open.

Kareem Williams — Founder. A product-delivery and engineering leader with 20+ years building the trading, risk, and regulatory platforms behind capital markets — cash equities, swaps, FX, rates and credit derivatives — at firms including Morgan Stanley, UBS, Citi, Credit Suisse, BNP Paribas, and Mizuho. He pairs delivery leadership with real technical fluency — he reads and writes the code (Java/Spring Boot, REST microservices, React, Azure) — and holds an MBA in Management of Technology (NJIT), an MS in Finance (Pace), and the PMP and PMI-ACP certifications. Blaque Baux turns two decades spent building the infrastructure of these markets on the strategies that run on them. Full profile at carterwarrens.com →

How to read, reproduce, and apply the research.