We extract investment signals from unusual data

iceberg

Information does not come in a pre-defined shape

Nicely structured time series are just the tip of the iceberg. Information relevant to your investment decisions can also come in text format, or even as image or video.

To make the best of these alternative sources is demanding. It requires state-of-the-art models and a clear sense of purpose.

AMS

retrieves niche investment signals from unstructured text data, for instance news websites, specialized newspapers or blogs. We build algos which scan through the web and digest the news the way an experienced analyst would, scaling up his skills and knowledge almost to infinity.

newspapers
teaching

Machine teaching

Our analysts teach the machine to recognize certain patterns relevant to a specific investment process or task (we implement mostly "supervised machine learning").

Our services

Signals we have created and optimized. We provide model-based selection of relevant news, as well as quantification of volumes and trends associated with certain themes.

For instance, our Inflation NewsBot routinely detects and analyses the news relevant to the near-term inflation forecast, from hundreds of thousands news sources, in multiple languages. Similarly, we detect signals relevant to specific economic sectors outlook (commodities, freight and semi-conductors).

Our data can be delivered straight to your inbox or programmatically accessed through an API, whichever works for you.

We can help you design your own models and fit them in your workflow, deploying state-of-the-art models at a fraction of in-house development cost.

The framework we have developed to extract signal from text can be replicated to almost any investment topic.

We can reestimate our models to fit your specific investment focus. We can also help you deploy the whole architecture that will allow you to build and develop your own models. Our understanding of the investment process is the key to developing tools you can actually make the most of.

About

Laurent Bilke

Laurent Bilke

CEO and Head of Research

Twenty years of experience in central banking and investment management, has driven the research effort in leading investment banks and hedge funds

State-of-the-art

We use mostly Transformers models, the latest advances in Natural Language Processing

With purpose

We don't just do models for their own sake, we know what investment management involves

Tailor-made

We can build signals relevant to your specific investment process

Featured from our blog