Signal Intelligence · Co-Investment Partnerships
Markets are networks.
We model them that way.
RavenGraph builds temporal graph neural networks that capture how information flows through financial markets — relationships, timing, and structure that time-series models miss entirely.
We partner with emerging hedge funds to put these signals to work.
You trade. We earn when you profit.
Most quant signals are built in isolation.
Markets aren't.
Traditional models treat assets as independent time series. But equities, commodities, and crypto don't move in a vacuum — they move through networks of relationships: supply chains, capital flows, correlated exposures, and sentiment contagion.
Temporal graph neural networks capture this structure. They model which nodes influence which, when, and how strongly. The result: signals that see around corners.
Aligned incentives.
No subscriptions. No retainers.
We earn only when our signals generate profits for your fund.
You deploy capital
Your fund allocates capital to a strategy powered by our graph predictions.
You trade our signals
Our graph models generate signals across equities, crypto, or commodities at flexible granularities — delivered in your preferred format.
We share in the upside
Flexible performance fee based on net profits. No payment if there are no profits.
→ Currently onboarding 5–7 partner funds across strategies.
Alpha beyond the backtest.
Real alpha survives execution costs, market impact, and regime shifts. Our partnership model aligns our incentives with your net P&L, not your data budget.
We deploy novel graph-based signals directly into your execution environment. It's not just a data feed; it's a co-managed strategy component built to perform.
Uncorrelated Alpha
Deploy signals derived from graph topology—relationship breakdowns, contagion, and flow—that provide orthogonality to standard momentum or mean-reversion factors.
Execution Loops
Data vendors deliver csvs and disappear. We optimize for your fills. Feedback from live execution drives our continuous model retraining and parameter tuning.
Asymmetric Upside
No management fees. We only monetize net new P&L generated by our signals. A pure performance partnership.
Temporal graph networks.
Not another LSTM.
Most quant shops run variations of the same models — LSTM, transformer, gradient boosting on price features. The signals degrade as they get crowded.
RavenGraph models the market as a dynamic graph: assets as nodes, relationships as edges, updated at each timestep. The network learns which relationships matter, when they break, and what that signals about price movement.
This is not an incremental improvement. It's a different class of model.
| Model type | Temporal Graph Neural Network (T-GNN) |
| Input structure | Multi-asset relational graph, flexible granularities |
| Signal horizon | Flexible · Minute to Daily |
| Coverage | Equities · Crypto · Commodities |
| Status | Live partner testing · Q1 2026 |
We are selective.
So are the funds we work with.
— Ideal Partner Profile
- Emerging fund, $50M–$500M AUM
- Strategy fit: equities, crypto, or commodities
- Execution infrastructure already in place
- PM open to signal integration (discretionary or systematic)
- Motivated by differentiation, not cost savings
— What We Bring
- Graph-based alpha signals, flexible granularities
- Strategy-specific model calibration
- Ongoing research and signal refinement
- Co-investment structure, performance-aligned
- Shared upside: flexible performance fee
We're speaking with a small number of fund managers this quarter.
If you run an emerging fund and are looking for a structural edge in your strategy — not a data subscription, a real partnership — we'd like to hear from you.