Research & Development

Our R&D branch

SILEX Technologies was incubated within the Ecole Polytechnique. Here our engineers and computer scientists develop sophisticated asset-management algorithms and the suite of programs that make up our SPARK platform

Technology R&D
PhDs, quant engineers, and data scientists
Led by Frank Génot (INRIA, MIT) and Fabrice Rey (X)
107 bytes of data processed daily
Machine
learning
R-Shiny : interactive
online platform
for data analysis
Constrained
optimization
Clustering /
stable replication /
Min Variance /
VaR calculations
Technology R&D
PhDs, quant engineers, and data scientists
Led by Frank Génot (INRIA, MIT) and Fabrice Rey (X)
107 bytes of data processed daily
Machine
learning
R-Shiny : interactive
online platform
for data analysis
Constrained
optimization
Clustering /
stable replication /
Min Variance /
VaR calculations

Most concepts of modern science are simple and can be expressed in language that is easy to understand

SPARK

Our technology

Our SPARK platform runs on a proprietary database and algorithms that are used selectively by our experts according to customized portfolio-management criteria. That allows us to incorporate our own investment convictions and the convictions of our clients – while keeping risk to a minimum

The core of our technology: the database

Houses an exceptionally large volume of data
(several million bytes collected daily)
Robust, cleaned and monitored daily
Houses an exceptionally large volume of data
(several million bytes collected daily)
Robust, cleaned and monitored daily
A golden copy
(data checks and corrections performed)
A golden copy
(data checks and corrections performed)

“Quantamental” analysis
a combination of fundamental research and quantitative analysis

analysts recommendation
financial strength
earnings quality
dividend yield

Sophisticated algorithms for efficient portfolio optimization

The main objective of our platform is to offer portfolios that generate an optimal compromise between expected return and risk, as stable as possible over time. For this purpose, SPARK exploits the latest advances in Operational Research, such as non-linear or even non-differentiable optimization, in high dimension, the regularized multilinear regression applied to time series or the estimation of the correlation between asynchronous time series