Arabesque’s AI Engine aims to forecast the future returns of 25,000 equities daily

Powering AutoCIO is Arabesque’s AI Engine, a massively distributed computational graph consisting of 255 million nodes and 750 million edges. The AI Engine ingests raw data, applies machine learning algorithms, and outputs a probability weighted expected return daily for each stock.

Request a demo

ai engine

Arabesque AI – by numbers

arabesque ai by numbers

Why do we use AI in
Portfolio Management?

Data today is produced in unprecedented quantities, increasing the need for intelligent machines to discover data relationships that can be translated into alpha opportunities. At Arabesque AI, we predict stock performance and other features by utilising information from big and diverse datasets, at a scale humans simply could not replicate. This enables us to draw out new sources of alpha.

ai
why use ai

The AI Engine simulates a traditional investment approach on an enormous scale

AI in practice

Our process

- Forecasting stock performance - universe of 25,000 stocks

- Applying an AI model to process big data using deep learning and other machine learning techniques

- The data - inputs and outputs - is constantly changing - the AI engine is continually learning, retraining, imporving and adapting the mapping on 20+ years data

big data

Using AI

- Enormous set of data and mapping

- Constantly learns and adjusts the model

- No belief system/adaptive

alpha sources

Performance

- Generality across a wide universe of 25,000 stocks

- We get it right 52.7% of the time

- $400m investment strategies powered by Arabesque AI

Delivery

- Cost-effective customised active alpha investment strategies

- Allowing client-specific calibration of investment and sustainability preferences

- Cloud-navtive system - computations are seamlessly spread across multiple data centres in Europe

high Scalabality

Contact Us

For any questions regarding Arabesque AI and our investment products, or to talk to one of our advisors, please contact [email protected]