An independent practice spanning quantitative finance, AI systems architecture, and applied machine learning, building tools that stress-test claims, strategies, and systems before they cost real money.
Patterson Research is an independent research and engineering practice founded by Charles Patterson. The work spans two domains connected by a shared methodology: treat every claim (whether a trading strategy, a model's performance, or a system's reliability) as guilty until it survives rigorous testing.
Quantitative research. Applying deflation-adjusted, power-aware statistical testing to trading strategies and market anomalies. The core insight is that most published alpha is the product of multiple testing and parameter optimization. The remedy is falsification-first methodology: deflated Sharpe ratios, parameter-fragility gates, and confirmation holdouts that keep discovery honest.
AI engineering. Designing and building production AI systems: agent architectures, evaluation pipelines, and decision-support tools that hold models to the same standard. Not demos. Systems that work because they've been tested the same way you'd test a trading strategy, adversarially, with the failures counted honestly.
An open-source quant research engine with a falsification core. Combines deflated Sharpe ratio testing, parameter-fragility analysis, power-aware verdict taxonomy, and anti-mining deflation into a single pipeline. Built to kill bad strategies before they cost real money.
The qualitative counterpart to Penrose. Applies falsification principles to investment theses and narrative research, stress-testing qualitative claims with the same rigor the engine applies to quantitative ones.
Active investigation into systematic mispricing in Kalshi weather tail markets. The favorite-longshot bias in temperature event contracts exhibits a quantifiable, fadeable structure that survives deflation-adjusted testing. Full research note forthcoming.
AI engineering and strategy. Available for select consulting engagements: agent architecture, evaluation pipelines, and AI adoption strategy for teams that don't want to burn budget on approaches that don't survive scrutiny.
If you have a system that needs to work reliably or a claim that needs to be tested honestly, get in touch.