Launching today
Lumina Capital
Autonomous Multi-Agent AI Hedge Fund Studio
1 follower
Autonomous Multi-Agent AI Hedge Fund Studio
1 follower
Lumina Capital is an autonomous multi-agent quantitative hedge fund engine bridging academic AI and real-world execution. Features native Microsoft Qlib Alpha158 factor math, Google TimesFM probabilistic fan charts, LangGraph anti-sycophancy cross-debates with a CRO veto loop, and a 3D causal knowledge graph. 100% open source.
Hi Product Hunt! 👋 I'm the creator of Lumina Capital.
Over the past few months, I saw massive excitement around open-source AI hedge funds, but when I looked under the hood of existing implementations, they suffered from critical flaws: feeding raw price numbers directly into LLMs (causing hallucinations), ignoring market slippage, and having zero defense mechanisms during flash crashes.
I built Lumina Capital to solve this from an engineering perspective:
- 🧠 Multi-Agent Council: 7 legendary investor personas (Buffett, Burry, Soros, Cathie Wood, etc.) with anti-sycophancy adversarial debates and an absolute CRO risk veto.
- ⚡ Quantitative Math: Pure Python vectorized Qlib Alpha158 + LightGBM meta-decision filter.
- 📈 Probabilistic Forecasting: Google TimesFM 5-day fan charts with crash probability.
- 🛡️ 10-Second Sentinel: Physical dynamic ATR stop-loss floor that doesn't wait for LLM loops.
It's completely free and open-source under CC BY-NC 4.0 for research and simulation!
I'd love to hear your thoughts, feedback, or any feature ideas. Thank you so much for your support! 🚀