Cromwell develops large‑scale transformer models that analyze world news, fundamentals, and price time series to assess predictive signals for financial markets. Their flagship research project trains a 300‑million‑parameter model on live data and evaluates performance on a public $100 K paper portfolio, aiming to determine whether news can provide a measurable market edge.
Funding
Funding not disclosed
Founders
Product
Problem
Cromwell addresses the difficulty of extracting actionable investment signals from the massive, unstructured flow of global news, corporate fundamentals, and market price data. Traditional analysis methods struggle to integrate these heterogeneous sources in real time, limiting the ability to identify predictive market lift.
Solution
Cromwell is developing a 300‑million‑parameter transformer model that simultaneously processes world news articles, financial fundamentals, and price time series to evaluate their combined predictive power for market movements. The model is currently being tested live on a public $100 K paper portfolio, allowing continuous validation of its forecasting capability. By leveraging large‑scale deep learning, Cromwell aims to quantify the incremental lift that news and fundamentals provide over price‑only signals, delivering a data‑driven foundation for investment strategies. Research findings are slated for release in 2026, positioning the platform as a rigorous, evidence‑based tool for signal generation in quantitative finance.
Target Audience
Primary users are quantitative research teams, hedge funds, and asset managers seeking advanced AI‑driven signals to enhance their systematic trading models.
Features
- 300M‑parameter transformer architecture designed to ingest and fuse textual news, structured fundamentals, and historical price series
- Real‑time inference on live market data, evaluated against a transparent $100 K paper portfolio
- End‑to‑end training pipeline that aligns unstructured news sentiment with quantitative financial outcomes
- Continuous performance monitoring and hypothesis testing to assess predictive lift over baseline models
- Planned research deliverables and open results expected by 2026 for academic and industry scrutiny