Víctor Bustos@vitoac-22

We built the risk infrastructure of a quantitative hedge fund, not a retail day-trading bot. Our autonomous agent operates strictly in the derivatives market, bridging advanced stochastic modeling with AI-driven risk auditing via the Model Context Protocol (MCP). The pipeline fits a GARCH(1,1) model with skewed-t residuals to daily log returns, forecasting conditional variance to dynamically select volatility regimes (structuring Iron Condors for compression, or Straddles for VaR breaches). Risk is strictly managed through 12 deterministic Risk Gates that evaluate bid-ask spreads, liquidity, delta neutrality, and concentration caps before any capital is deployed. We do not let AI hallucinate strikes or position sizing. The quantitative engine calculates the exact atomic multi-leg (MLEG) JSON payload. This mathematical rationale is then sent to our Executive Risk Committee (Featherless Qwen-32B LLM). The LLM acts solely as a Portfolio Manager: it audits the logic and, if approved, triggers execution exclusively through a custom MCP server, strictly isolating Alpaca broker credentials behind the protocol standard.

🌎 Online Hackathon | 💻 Dates: 28 August–4 September 2026 | 🏆 $6,000 Prize Pool Build AI trading agents on Alpaca — autonomous agents and trading apps using Alpaca's Trading API, MCP server and CLI. 🌍 Fully online hackathon. Join and build from anywhere in the world.