Flying Dutchman Labs

Flying Dutchman Labs was born from the frustration of seeing sophisticated narratives wrapped around fragile portfolios. The lab is an independent R&D initiative focused on portfolio construction first, with investment ideas treated as inputs into a robust optimization and risk framework rather than standalone bets.

We develop both fully systematic strategies and hybrid, due-diligence-driven approaches, with a particular emphasis on short-selling weak or structurally flawed businesses. Systematic engines, agentic LLMs operating on newsflow and company disclosures, and bespoke risk models all plug into the same portfolio construction and risk management stack.

Flying Dutchman Labs is not an asset manager, it is a research workshop. The outputs are strategies, tools, and engines that can be embedded into institutional investment processes via API, custom research, or white-label solutions.

Our work focuses portfolio construction, not stock picking. We design allocation engines using convex optimization and multi-objective methods (NSGA-III, MOEA/D and related algorithms) to search directly over the Pareto frontier of risk, return, drawdown, and factor exposure. Instead of treating risk as “just volatility,” we integrate tail metrics, scenario losses, and structural constraints (liquidity, borrow, turnover) into the optimization problem itself. The result is a portfolio that is engineered to be coherent with the investor's true risk budget rather than reverse-engineered from ad-hoc trades.

We develop systematic volatility engines that plug into the portfolio core as both alpha and protection sleeves. In benign regimes, they exploit relative dislocations across single-stock options, index volatility, and term structures; when regimes deteriorate, the same framework shifts toward defensive positioning, emphasizing convexity, drawdown control, and chaos protection. Positioning, hedging, and leverage remain governed by the portfolio-level optimizer and risk engine, so every volatility trade is sized not only for standalone opportunity but for its contribution to total portfolio resilience.

Our equity L/S work is deliberately hybrid: argument-driven longs and forensic shorts. On the long side, we use agentic LLMs (e.g. LLM-based research agents that autonomously read newsflow, filings, and other primary documents) to build and update the qualitative argument behind a position. On the short side, we spend disproportionate time on deep due-diligence: accounting quality, governance, capital structure, business model fragility, and any gap between what companies say and what they actually deliver. We enjoy “punishing” practices that are misleading to investors by translating those red flags into high-conviction, risk-aware short positions.

We believe every portfolio manager is first a risk manager, so the lab is built around coherent risk measurement and robust simulation engines. Portfolios are evaluated with a battery of metrics (drawdowns, tail measures, factor and contagion risk) under bootstrapped and Monte Carlo scenarios rather than a single historical backtest. We also use deep generative models such as TimeGANs and diffusion-style scenario generators to create alternative histories and stress regimes, helping to test whether a strategy's edge survives outside the sample it was born in. This architecture is designed to continuously challenge our own ideas before the market does.