Idea
Deep learning model delivering robust, risk-adjusted macro portfolio returns across diverse market regimes.
Research Paper
Core Innovation
This paper introduces DeePM, which uniquely combines a Directed Delay mechanism to handle asynchronous data, a Macroeconomic Graph Prior to regularize cross-asset dependencies, and a distributionally robust objective using a smooth worst-window penalty as a proxy for Entropic Value-at-Risk. These innovations enable superior risk-adjusted performance and regime robustness compared to prior trend-following and transformer-based models.
Why It Matters
Systematic macro portfolio management faces challenges from asynchronous data, noisy signals, and volatile market regimes. DeePM improves risk-adjusted returns and resilience, enabling asset managers to better navigate adverse market conditions and regime shifts. This scalability and robustness can transform portfolio management workflows and enhance investment performance.
Market Size (TAM)
$20–50B TAM for quantitative asset management platforms; $2–10B SAM from hedge funds and institutional investors. Driven by increasing demand for robust, AI-driven portfolio management and risk mitigation.
Potential Customers & Pain Points
- Asset managers – Need robust portfolio strategies resilient to regime shifts
- Hedge funds – Require improved risk-adjusted returns under transaction costs
- Quantitative traders – Seek models handling asynchronous and noisy financial data
- Institutional investors – Demand consistent performance across volatile markets
Business Model
Subscription-based SaaS platform offering access to DeePM models and analytics; licensing to hedge funds and asset managers; consulting for integration and customization.
Competitive Landscape
- AQR Capital Management
- Two Sigma
- Renaissance Technologies
- Man AHL
- Momentum Transformer
Implementation Challenges
- Integration with existing portfolio management systems
- Regulatory compliance and transparency requirements
- Market adoption inertia for new AI-driven strategies
- Data quality and latency constraints in live trading
Validation Strategy
- Conduct live trading pilots with partner hedge funds
- Benchmark performance against existing portfolio strategies in real-time
- Perform stress tests across multiple market regimes
- Gather user feedback for iterative model and platform improvements
Research Paper Overview
DeePM: Regime-Robust Deep Learning for Systematic Macro Portfolio Management
Summary
DeePM is a deep-learning macro portfolio manager designed to maximize robust, risk-adjusted returns by addressing asynchronous data, low signal-to-noise ratios, and distributional robustness. It outperforms classical and state-of-the-art strategies in large-scale backtests across diverse market regimes using only daily closing prices.