Goal: Build an open-source, real-time signal intelligence system that detects early anomalies across news, prediction markets, shipping flows, and geo-economic data — inspired by Bloomberg Terminal, but open, modular, and research-driven.
Financial markets move before prices move.
This project aims to detect early, non-obvious signals hidden in:
- News sentiment shifts
- Prediction market odds (election, policy, conflict)
- Shipping & oil tanker movement
- Geo-statistical anomalies
- Macro narrative changes
The platform aggregates heterogeneous data sources, applies statistical + ML-based anomaly detection, and surfaces actionable signals before they appear in traditional market indicators.
- Markets are driven by information diffusion, not just fundamentals
- Early signals appear in alternative data before price action
- Cross-domain correlation (news + shipping + bets) increases signal strength
- Open research tooling beats closed black-box indicators
- Reddit (finance, geopolitics, macro subreddits)
- News APIs (global & regional)
- Social narrative velocity & sentiment drift
- Kalshi (macro, policy, elections)
- Betting odds as probabilistic signals
- Sudden probability regime changes
- Oil tanker movement & port congestion
- AIS-derived trade flow anomalies
- Energy logistics stress signals
- Trade disruptions
- Regional instability indicators
- Macro supply-chain stress metrics
- Statistical anomaly detection (Z-score, EWMA, regime shifts)
- NLP-based sentiment & topic modeling
- Time-series change-point detection
- Cross-signal correlation scoring
- Noise filtering & false-positive suppression
- Early-warning alerts
- Confidence-weighted signal scores
- Visual dashboards (research-focused)
Backend
- Python
- Pandas / NumPy
- FastAPI
- PostgreSQL / DuckDB
ML / NLP
- Transformers (FinBERT / domain-adapted models)
- Topic modeling
- Time-series ML
Infra
- Modular pipelines
- Offline-first design
- Reproducible research notebooks
Frontend (Later Stage)
- Research dashboards
- Signal timelines
- Correlation heatmaps
- Detect geopolitical risk before equity drawdowns
- Identify energy supply shocks before oil price spikes
- Track election-related macro volatility early
- Monitor narrative-driven market regime shifts
- Research & architecture planning
- Data ingestion pipelines
- Sentiment & anomaly models
- Cross-signal scoring engine
- Dashboard & visualization
- Public beta release
Target live version: End of year
This project is:
- Research-first
- Transparent
- Modular
- Designed for extensibility
Contributions, critiques, and research discussions are welcome.
This project is for research and educational purposes only.
It does not provide financial advice or trading recommendations.
Manthan Kumar
Research-focused developer working at the intersection of:
- Quantitative finance
- Alternative data
- Signal intelligence
- Applied machine learning
Most market tools react.
This platform is built to anticipate.