"LOQ team on Alpaca AI Trading Agents Hackathon Hackathon"

Team Idea

An autonomous options-trading agent for Alpaca paper trading, built around one rule: the LLM proposes, deterministic Python decides. Every order passes a statistical validation gate and a hard-coded risk layer before it can reach the broker. We measured what happens without that separation. Across 240 replays at temperature 0 with byte-identical inputs, 100% of the model's decision turns changed on replay — 40 of 40. A live third-party arbiter approved a $172,000 order against an $8,000 cap; the deterministic gate refused it. An 18-attack adversarial harness reached the broker zero times. The gate then refused all 21 of our own strategy/symbol candidates. We shipped that result rather than loosening the gate.

Team Members

PANNALA SAI RISHWANTH REDDY

pannala_sai_rishwanth_reddy482

Drona Venkata Abhiram

drona_venkata_abhiram873

Submission

Alpaca Options Agent

An autonomous options agent on Alpaca where the LLM proposes and deterministic Python decides. We measured our own model: 100% of its decision turns changed on replay at temperature 0. Then our validation gate refused all 21 of our strategies.