Investments

Our investment stack is an end-to-end agentic pipeline. Ingestion: 10-Ks and 8-Ks, sell-side notes, satellite imagery, anonymized card flows, Level-2 order books. Reasoning: Bayesian belief-updating across heterogeneous signals, causal-inference to separate idiosyncratic alpha from regime drift, Monte-Carlo path simulations to size positions against tail risk. Execution: a reinforcement-learning router that minimizes market-impact slippage. Every decision is logged with provenance; every model output is back-testable; every dollar of P&L is attributable. The desk is smaller, the coverage is wider, the cost-per-decision falls by an order of magnitude — and the edge compounds the longer the system runs.