We build general-purpose AI models that use first-principle reasoning to understand and anticipate market behaviour, delivering directional insight across asset classes, timeframes, and regimes.
Statistical models fail when market conditions differ from their training data. We build AI systems that reason through first principles to identify directional market trends.
Traditional AI approaches in finance lack a true understanding of market behaviour, needing constant retraining as markets change. Our foundation models provide a generalised understanding of market dynamics, that autonomously adapt to new market regimes without retraining.
We believe price movements originate in high-frequency data and propagate to lower timeframes. Our unified model architecture reasons across this hierarchy, delivering a holistic view of market behaviour.
Navigating constantly changing financial markets requires models that can generalise beyond historical data. Our key breakthrough is an end-to-end model training technique that learns directly from market experience, without reliance on pre-set rules. This capability enables our models to reason from first principles, allowing them to understand and adapt to evolving market environments, rather than just following patterns learned from the past.
Our proprietary architecture combines category theory (to understand market representations), deep learning (to identify generalisable representations), and symbolic AI (to embed reasoning). This moves beyond mere pattern-matching to extract structure and meaning from raw price data, turning noise into actionable insight.
We believe modelling market behaviour and creating a trading strategy are different from an objective, output and evaluation perspective. The primary goal of modelling market behaviour is to understand and explain how and why markets move, while the goal of a trading strategy is to generate profit by exploiting market inefficiencies or patterns.
While creating a trading strategy is about defining actionable rules for profit generation, the ultimate success of any strategy relies on the depth and adaptability of its underlying insights into market behaviour. Our core offering of generalised market understanding is a solution to the most pressing problem affecting predictive models in finance.
By developing models that leverage first-principle reasoning to anticipate market movements and navigate real-time volatility, we have achieved a level of market understanding that statistical pattern recognition models do not have. This foundational understanding is invaluable because it provides the critical context that many models lack. This context isn't limited to optimising a single trading strategy; its application is far larger, informing a wide range of critical functions across the financial markets, from investment decisions and risk management to in-depth market research.
Trading strategies are at the application layer; a truly intelligent model belongs on the foundation layer, driving improved decision-making in complex markets.
Real-Time Insights Driven by Market Understanding.
Our latest model identifies directional trends by predicting how likely price action in the selected timeframe will propagate to lower-frequency timeframes, distinguishing structurally significant market moves from noise.
The fundamental challenge for statistical AI models in finance lies in their dependence on historical patterns. This creates a critical vulnerability during market regime changes. When market shifts occur, these models struggle to generalise outside of their training data, resulting in inaccurate predictions and the need for constant retraining.
Our single pre-trained model provides actionable directional price insights with unprecedented generalisation across all asset classes, timeframes, and market regimes. This eliminates the complexity of managing multiple specialised models and the need for constant retraining.
Our foundation model uses end-to-end learning to build a fundamental understanding of market dynamics. This training generates transferable knowledge, creating insights grounded in first principles that instantly apply across any asset class, timeframe, or challenging market regime.
Unlike traditional approaches that search for fixed patterns, our model analyses real-time market data at inference by applying its understanding of market behaviour. This deep understanding of market behaviour is supported by the model's unified architecture, which facilitates explainability by tracing insights to their source and strengthens them through alignment across multiple timeframes.
Unlike models with fixed rules that degrade over time, our model has no fixed features, making it inherently resistant to concept and feature drift. This critical advantage ensures it consistently delivers accurate directional insights and maintains predictive power, providing reliability in volatile markets without the need for market specific training data or constant retraining.