Hi, I'm Davide. Beleth is my submission for the Alpaca AI Trading Agents Hackathon, a project built from scratch in a few days, with the stated goal of winning, not just showing up.
The work looked more like risk engineering than trading: what took most of my time wasn't "getting the agent to trade," it was building the gates that stop it from making bad trades; thresholds calibrated on real historical data (VIX 1990-2026), sizing rules kept properly separate from each other, an aggregate cap on committed risk, and a deploy discipline written into the code itself (a script that blocks redeploys while the market is open, not just a rule on paper).
After the first day of live trading I ran a full architectural review of the risk system instead of applying isolated patches, because the quality of the risk system matters more than the number of trades executed.
I'm on this forum to answer questions, discuss design choices and collect feedback. If anything about Beleth's behavior looks strange or wrong to you, tell me in Strategy & Signals or Site Feedback, that's exactly the kind of signal I want.
About me:
27 y.o. from Italy, Full Stack Developer, videogamer, passionate about cats, japan, medieval armors, AI and cars.
My favorite trading ticker:
-
Find more about me on my website, davidemaiorana.dev, or connect with me on LinkedIn here.