🌲 Evergreen Capital - MVP
- Solana Track 1st Place
- Polymarket Track 2nd Place
Project Not Completed
- Under Construction Until the legalization of Poly Market Betting In the USA - (Bare Bones MVP)
- Mock Data
We wanted to answer a simple question:
- What would a hedge fund look like if it were run entirely by AI LLMS and executed entirely on-chain?
Prediction markets already encode the world’s information, and Solana gives the speed to trade them. We built Evergreen to combine:
- Institutional-style research
- Multi-agent debate
- Fully automated execution
- On-chain portfolio management
- A hedge fund that never sleeps.
Evergreen allows anyone to stake SOL into a shared liquidity pool managed by a network of Gemini research agents.
The system:
- Scrapes Polymarket markets + external web data
- Loads everything into Snowflake, our data warehouse
- Sends structured data to five specialized Gemini analysts
- Analysts independently research → argue → vote
- A consensus vote triggers autonomous trade execution on Solana
- Evergreen tracks open/closed positions and distributes returns to pool contributors
Essentially:
- You deposit SOL → the AI hedge fund trades → profits return to you.
Evergreen uses five Gemini agents, each modeled after a different hedge-fund analyst archetype:
- The Quant – statistical edge, probability weighting
- The Macro Analyst – event flow, catalysts, sentiment
- The Skeptic – risk management, counter-arguments
- The Data Miner – anomaly detection, micro-signals
- The Trader – execution timing and conviction scoring
Agents debate the trade just like a real investment committee. But faster.
- Data Pipeline & Snowflake Warehouse
- Scrape Polymarket markets + external narratives
- store in Snowflake
- Feature-engineer time-series for agent consumption
- AI Research & Debate Engine
- Gemini agents receive identical data packets
- Agents independently produce theses
- Structured debate system forces cross-examination
- Weighted voting produces a final trade signal
- Solana Execution Layer
- Handles deposits + withdrawals
- Opens/settles Polymarket positions
- Tracks NAV, Balance, exposure, open trades
Distributes returns back to participants
- AI hedge funds live or die by data quality
- Multi-agent research dramatically improves conviction
- Solana is the ideal execution layer for autonomous trading
- Prediction markets reward disciplined, systematic strategies
- Vibe Coding is supper hard at 2:00 am
- Testing this in a 24hr period is hard, polymarket is ilegal in USA
- Polymarket becomes legal in the USA this year (2025), Until then we will be working on other consulting projects.
- integrate full back end and full front end, data now is most mock data.
- Expand our data lake: social sentiment, news vectors, on-chain flows
- Add risk models, portfolio constraints, and VaR-style limits
- Support additional prediction platforms
- Move toward a fully permissionless “stake → earn” model
- Evergreen aims to become the first fully autonomous, transparent, on-chain hedge fund powered by AI.
- Backend: FastAPI, Python, Pydantic, Uvicorn
- AI Research Desk: Gemini
- Blockchain: Solana, WalletConnect
- Data Layer: Snowflake, SQL
- Frontend: React, TypeScript, Tailwind, Vite, Recharts
- Execution & Logic: Rust, Polymarket