TUS Capital runs delta-neutral strategies in digital-asset arbitrage and volatility. Our edge is a research program: language-model systems for signal analysis, research automation, and execution.
We are a systematic, market-neutral fund. Positions are hedged; returns are engineered from structural inefficiency, not directional conviction.
Since 2022, the fund has operated an internal applied-research program treating large language models as production infrastructure — subject to the same evaluation, monitoring, and failure-mode discipline as any trading system. Findings from this work are published openly.
The team works from Toronto, Canada — a small group of researchers and engineers drawn from quantitative trading, distributed systems, and machine-learning research. We keep the firm deliberately small: every strategy in production has an owner who can explain, line by line, why it makes money and how it fails.
Digital-asset basis, funding, and volatility structures, traded with hedged exposure and continuous collateral-risk monitoring. The book is constructed so that no single venue, instrument, or funding regime dominates risk.
Agentic pipelines connect language models to live market-data feeds for signal analysis and research automation, with experimental findings translated directly into deployed tooling. Models propose; deterministic systems and humans dispose.
Reproducible evaluation frameworks, failure-mode analysis, and governance patterns for tool-using financial agents — before any system touches capital. Every automated decision path has a kill switch, an audit trail, and a documented worst case.
Selected notes from the fund's applied-research program on AI reliability and governance.
// Full texts available on request.
Toronto, Canada. We respond to research correspondence and introductions; we do not accept unsolicited subscription requests.
INVESTOR LOGIN →