TradeSmith
Fintech, Investment Analytics

TradeSmith

TradeSmith integrates Nixtla’s TimeGPT to forecast 22,000+ financial time series daily—delivering half-a-million forecasts per month with minimal tuning.
TradeSmith

22,000+ Daily Forecasts

Forecasted time series every day, powering 100,000+ forecasts per month.

Improved Model Performance

TimeGPT surpassed legacy xgboost models in both accuracy and speed with minimal tuning.

Reduced Complexity Simplified Feature Engineering

Cut the number of required exogenous variables and streamlined integration of historic and future features.

About TradeSmith

Overview

TradeSmith is a financial technology company delivering investment research, portfolio analytics, and predictive tools to over 60,000 active investors.

The platform uses Nixtla’s TimeGPT to forecast more than 100,000 forecasts each month.

By replacing tree-based models in production, TradeSmith achieved higher accuracy, faster predictions, and greater scalability across their analytics suite.

With support from Nixtla’s scientific team, they tailored TimeGPT through professional services to meet their unique investment forecasting needs.

TradeSmith overview

60,000+

Active Investors

100,000+

Monthly Forecasts

22,000+

Time Series Forecasted
BUSINESS CHALLENGES

The Challenge

TradeSmith needed to scale forecasting to tens of thousands of time series daily, improve accuracy over legacy models, and simplify the integration of exogenous features.

  • High complexity and volatility of financial time series forecasting.
  • Legacy tree-based models struggled with accuracy, speed, and scalability.
  • Feature engineering was cumbersome due to numerous exogenous variables.
OUR APPROACH

The Solution

Nixtla’s TimeGPT was integrated into TradeSmith’s production pipeline, replacing tree-based models and leveraging zero-shot capabilities to deliver reliable forecasts at scale.

  • Forecasted 22,000+ series daily with zero-shot predictions.
  • Outperformed xgboost models in accuracy and inference speed.
  • Streamlined exogenous variable handling with minimal tuning.
Michael Carr.'s company logo

TimeGPT outperformed our legacy xgboost models, delivering more reliable forecasts with minimal tuning and scaling effortlessly to half-a-million forecasts per month.

Michael Carr.
Chief, Quantitative Research
TimeGPT drove superior forecast accuracy, massive scale, and simplified feature workflows—empowering TradeSmith’s investors with actionable insights.

Business Outcomes

High-Volume Forecasting

  • 22,000+ daily time series forecasts
  • 100,000+ monthly forecasts

Improved Accuracy & Speed

  • Superior forecast accuracy
  • Faster prediction times

Streamlined Analytics

  • Easily add historic and future exogenous variables
  • Lower maintenance overhead