Choose your location
Europe
Asia
America
Africa

Hightech

Consulting and Big Data optimisation on AWS for fraud prediction

OUR CUSTOMERS

Our client is an expert in the management and simplification of recurring payments.

+ 0 YEARS
experience in managing and simplifying payments
0 M €
of funds raised
0 employees
+ 0 années
d'expérience dans la gestion et la simplification des paiements
0 M €
de levée de fonds réalisée
0 collaborateurs

CONTEXT & ISSUES

The migration from Hortonworks to an Amazon EMR cluster was accompanied by a consultancy mission to optimise the solution.

Objectives:

  • Make full use of the CPU / RAM capacities made available by EMR
  • Enable scaling of the solution, and thus the construction of a Machine Learning model for fraud prediction

TEAMWORK'S RESPONSE

Optimise use of the EMR cluster to take advantage of its full capacity and be able to
scaling in response to future increases in data volumes.

Our approach is as follows.

  • Analysis of data typology (schema, format, etc.)
  • Analysis of the source code of the application running on the cluster (parallelism, shuffles, etc.)
  • Analysis of the EMR cluster configuration (applications installed, number of workers, memory allocation, spots, etc.)

The success of this phase is optimal use of the EMR cluster by observing 100% CPU across all nodes, sufficient memory allocation to prevent swapping and optimal data parallelism from Amazon S3.

Solutions: Amazon EMR, Amazon S3, Jupyter, Hadoop

THE BENEFITS

  • Ability to build and train a Machine Learning model
  • Effective implementation of a fraud prediction solution
  • Provision of applications via a REST API (python – Gunicorn server)

Discover our customer references

Contact us
x
* Required
x
Choose your location
USA Canada FR Canada EN France Luxembourg EN Luxembourg FR Germany Slovakia Bern/Zurich Geneva India China Vietnam Singapore Mauritius Mauritius Germany Suisse