Decentralized marketplace for financial AI models on Bittensor. Miners compete on price prediction, sentiment & risk . Validators score against real market data, best models earn $TAO.
AI Finance Oracle is a Bittensor subnet (#7) that creates a decentralized competitive marketplace for quantitative financial AI models.
Problem: Institutional-grade predictive models are locked behind hedge funds. Retail traders, DeFi protocols, and smaller funds lack access to state-of-the-art financial intelligence. Existing oracles only fetch historical data — they don't predict.
Solution: Miners compete to build the best AI models for price prediction (60%), sentiment analysis (25%), and risk assessment (15%). Validators verify every prediction against real market data from exchanges like Binance. The most accurate models earn $TAO rewards via Yuma Consensus.
Scoring: Predictions are scored on directional accuracy, magnitude accuracy (MAE), confidence calibration, and latency — with a 1.5x bonus for correct calls in volatile markets.
Key Features:
- 6 specialized AI miners (QuantFlow, AlphaNet, DeepSentiment, RiskForest, etc.)
- 3-4 validators with real-time market data verification
- Interactive web demo with 3 live scenarios (BTC prediction, ETH sentiment, portfolio risk)
- Full REST API with Swagger documentation (20+ endpoints)
- Yuma Consensus with TAO reward distribution
Target Market: DeFi protocols ($100B+ TVL), retail traders, algorithmic trading ($20B+ market), and quantitative researchers.
Tech Stack: Python, FastAPI, Pydantic, Bittensor SDK
Week 1 — Research & Design
- Analyzed existing Bittensor subnets (Taoshi/Prophet) for financial oracle patterns
- Designed miner/validator incentive mechanism with MAE-based scoring
- Wrote SUBNET_PROPOSAL.md with full mechanism specification
Week 2 — Backend Development
- Built FastAPI backend with 20+ API endpoints (miners, validators, network, demos)
- Implemented 3 miner task types: price prediction, sentiment analysis, risk assessment
- Created in-memory database with 8 pre-seeded miners and 3 validators
- Developed scoring engine with weighted multi-factor evaluation
Week 3 — Frontend & Demo
- Built interactive dark-theme web UI with 3 demo scenario cards
- Implemented real-time loading animation simulating subnet broadcast
- Created detailed result views showing miner responses, validator checks, and consensus
- Added Swagger API docs and ReDoc for full API exploration
Week 4 — Polish & Submission
- Tested all 3 demo scenarios end-to-end (BTC prediction, ETH sentiment, portfolio risk)
- Wrote pitch deck, demo video script, and visual guide
- Deployed to GitHub with comprehensive README and judge instructions
- Prepared demo video and pitch presentation
Currently bootstrapped — no external funding raised.
This project was built during the Bittensor Subnet Ideathon as a proof-of-concept. All development has been self-funded by the team.
Future plans:
- Apply for Bittensor subnet registration (requires TAO stake)
- Seek grants from Opentensor Foundation for subnet development
- Explore partnerships with DeFi protocols for integration funding
- Revenue model: Premium API access for institutional-grade predictions
We are open to conversations with investors and partners interested in decentralized financial AI infrastructure.