EIGEN-12 is an autonomous, multi-agent AI reasoning engine built for the institutional buy-side. We replace the traditional, high-overhead human research desk with a specialized AI swarm that ingests, debates, and executes on complex market hypotheses in seconds. Crucially, we integrate real-time satellite telemetry and climate risk data as a primary alternative data feed, allowing our system to price physical climate risks and supply chain disruptions long before traditional news outlets report them.

Inspiration

The institutional finance industry is facing a dual crisis: a "Compliance Wall" that blocks black-box AI adoption, and a massive "Alpha Gap" in processing unstructured data. As noted by quantitative leaders at top asset managers, traditional LLMs collapse under the complexity of cross-asset relationships. Furthermore, < 5% of available alternative data is utilized by human analysts. We were inspired by a specific blind spot: climate and geospatial risk. When a drought impacts a critical shipping canal or flooding disrupts a semiconductor supply chain, human analysts take days to synthesize the impact. We asked: What if a single founder could deploy an AI swarm that reads satellite imagery, debates the macroeconomic impact, and generates a fully auditable trade thesis in 12 seconds? EIGEN-12 is the answer. It is the ultimate One Person Company (OPC) thesis: infinite cognitive leverage with near-zero marginal cost.

What It Does

EIGEN-12 is not a chatbot; it is a real-time Cognitive Terminal. When a market trigger occurs, our system deploys a swarm of specialized agents: Satellite & Climate Agent: Ingests geospatial feeds (e.g., port congestion, crop health indices, flood mapping) to detect physical disruptions. Macro & Logistics Agents: Analyze the downstream economic impact of the climate event on specific equities or commodities. Quant Risk Agent: Stress-tests the hypothesis against historical volatility and portfolio constraints. PM Arbiter: Weighs the agents' conflicting arguments and outputs a final, 100% auditable chain-of-thought trade recommendation, complete with a cryptographic compliance export.

How We Built It

We architected and deployed this entire multi-agent system in a hyper-velocity 6-hour hackathon sprint, proving the extreme capital efficiency of the OPC model. Backend: Built a custom asynchronous orchestration layer using Python and FastAPI to manage stateful conversations between agents. AI Core: Utilized advanced prompt engineering and retrieval-augmented generation (RAG) to ground the agents in both financial 10-K data and simulated satellite/climate telemetry APIs. Frontend: Developed a low-latency, real-time streaming UI using Next.js and React to visualize the "dialectical debate" as it happens. The quantitative impact of this architecture is stark:

Challenges We Ran Into

The "Echo Chamber" Problem: Early versions of the swarm suffered from groupthink, where agents simply agreed with each other's hallucinations. We solved this by engineering a Dialectical Debate Framework, explicitly prompting agents to adopt adversarial roles (e.g., "Bull" vs. "Bear") and penalizing unsupported claims. The Compliance Wall: Judges and institutional clients rightly fear AI "black boxes." We had to pivot from simply outputting a trade to generating a structured, immutable JSON audit trail that maps every data citation to every logical step, making the AI's reasoning legally defensible. Integrating Unstructured Geospatial Data: Translating raw satellite/climate signals into actionable financial logic required careful context-window management to prevent the LLM from being overwhelmed by noisy telemetry data.

#What We Learned Building EIGEN-12 in 6 hours proved that the One Person Company (OPC) model is no longer a compromise; it is a competitive advantage. By leveraging AI coding assistants and autonomous agent architectures, a single founder can achieve the output of a $1.5M/year quant team. We also learned that the true moat in AI is not the data itself, but the interaction history of the agents. Every debate our swarm has generates a proprietary dataset of financial and climate reasoning, creating a compounding flywheel that simple RAG applications cannot replicate.

What's Next for EIGEN-12

Live Satellite Integration: Partnering with providers like Planet Labs or Sentinel Hub to ingest real-time, high-resolution geospatial climate data. Dialectical Fine-Tuning: Moving beyond prompt engineering to fine-tune the PM Arbiter model on our proprietary dataset of agent debates, teaching it to weigh conflicting financial and climate logic with institutional-grade accuracy. Enterprise FIX Protocol: Integrating with broker-dealer APIs to move from "recommendation" to autonomous, T-0 (instantaneous) trade execution.

Built With

• codex