The digital quant: instant portfolio optimization with JointFM

2026-02-16 · Source: Blog | DataRobot · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Data Science & Analytics, Emerging Technologies & Innovation · Depth: Advanced, medium

Summary

JointFM is presented as the first AI foundation model for zero-shot joint distributional forecasting in multivariate time-series systems, capable of generating coherent future scenarios in milliseconds. This model aims to overcome the limitations of traditional quantitative methods, which are rigid and slow, and existing time-series foundation models that lack cross-variable dependency capture. JointFM is pre-trained on an infinite stream of synthetic stochastic differential equations (SDEs), making it domain-agnostic and able to predict full joint probability distributions for systems like power grids or stock portfolios. The model demonstrates its utility in quantitative finance for instant portfolio optimization (IPO), replacing overnight batch processes with real-time rebalancing and adaptation to new market conditions without retraining. Benchmarking against classical Geometric Brownian Motion (GBM) shows JointFM achieves comparable risk-adjusted returns with significantly faster simulation times.

Key takeaway

For AI Scientists and Machine Learning Engineers building financial applications, JointFM offers a paradigm shift from traditional modeling to AI-driven solutions. You can now achieve real-time portfolio optimization and risk management by leveraging its zero-shot, multivariate forecasting capabilities. This eliminates the need for constant model retraining and complex pipelines, allowing for instant adaptation to market changes and integration with AI agents for ad-hoc business decisions.

Key insights

JointFM is a zero-shot foundation model predicting full multivariate distributions for instant, domain-agnostic time-series forecasting.

Principles

• Synthetic pre-training enables broad generalization.
• Joint distributions are critical for systemic risk assessment.
• Parallel decoding accelerates scenario generation.

Method

JointFM uses a transformer-based architecture with factored attention for high-dimensional context, heavy-tailed distributional heads for robust predictions, and parallel decoding for instant, simultaneous future horizon prediction.

In practice

• Integrate JointFM for real-time portfolio rebalancing.
• Use for instant decision-making in complex systems.
• Apply to diverse domains beyond finance.

Topics

• JointFM
• Zero-shot Forecasting
• Multivariate Time-Series
• Portfolio Optimization
• Foundation Models

Best for: Machine Learning Engineer, AI Scientist, AI Engineer, AI Data Scientist, Research Scientist

Related on AIssential

• JointFM-0.1: A Foundation Model for Multi-Target Joint Distributional Prediction — JointFM is a zero-shot foundation model predicting future joint probability distributions by pretraining on synthetic SDEs.
• TTM: Tiny Foundation Models for Multivariate Time-Series Forecasting — Tiny Time Mixers (TTMs) enable efficient, accurate time-series forecasting via compact pre-trained models and multi-level adaptation.
• Zero-shot Multivariate Time Series Forecasting Using Tabular Prior Fitted Networks — Multivariate time series forecasting can be reframed as scalar regression for zero-shot tabular foundation model application.
• Tyan-WP: A Wind Power Foundation Model for Ultra-Short-Term Probabilistic Forecasting — Tyan-WP is a foundation model for wind power forecasting, achieving accurate zero-shot predictions by integrating static site data and meteorological covariates.
• Five Questions About Chronos-2, the Time Series Foundation Model — Time series foundation models like Chronos-2 enable zero-shot forecasting by learning recurring temporal shapes from diverse data.
• A Multi-Agent Framework for Quantitative Finance : An Application to Portfolio Management Analytics — A Multi-Agent framework autonomously performs complex quantitative finance tasks using dynamic code generation and specialized agents.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Blog | DataRobot.