Data Science Intern - Lyon
Location:Lyon, Rhone-Alpes, France
Area of InterestInformation Technology
Technology InterestBig Data, Analytics
Start Date: March/April 2020
Duration: 6 months
Location: Lyon, France
Sentryo, a ~40 people startup specialized in industrial cybersecurity and based in Villeurbanne, is now part of Cisco’s IOT business unit. It develops a software, “Cyber Vision”, already used by many customers, that enables industrials to get a complete overview of the interactions occurring in their networks using deep packet inspection as well as ensure security and integrity of such systems.
Within Sentryo, data science functionalities for Cyber Vision are developed by a small team which you will join for this internship. Amongst over subjects, this team currently works on time series prediction to provide additional anomaly detection capabilities to Cyber Vision. One of the main challenges of this team is the unicity of the installation of each client. Indeed, this prevents the possibility to train “in house” a single algorithm that would work for every client and requires to design automated processes to find suitable models.
Currently, we mainly use neural networks as models for time series predictions because of their flexibility and performance. Nevertheless, these algorithms have lots of tweakable hyperparameters, in addition to their architectures that can also be seen as hyperparameters, that can influence their performance. This led us to being interested in AutoML (Neural Architecture Search, Hyperparameters Optimization, Auto Feature Engineering…), a recent field of research that has been rapidly growing in the last years, which could help us find suitable models with less human supervision.
During this internship, your mission will be to identify, then implement, state-of-the-art techniques to improve the process of finding relevant models, leveraging the capabilities offered by AutoML.
Several papers on AutoML, which could be used during this internship, can be found here :
What we are looking for a student finishing his M2+ studies in a data science or machine learning related specialization and available for a 6+ months internship.
In addition, the candidate should:
· be curious, technophile with a strong appeal for machine learning
· master one programming language (ideally Python)
· have a good background in the mathematics used in data science (statistics, linear algebra, …)
· have a solid background in machine learning, including some implementations experience
· ideally have some knowledge about neural networks/deep learning