CONTEXT

At Wiremind, the data-science team is responsible for the development, monitoring and evolution of all ML-powered forecasting and optimization algorithms in use in our Revenue Management systems through a library called Pythie. Our algorithms are divided in 2 parts:

• a modelisation of the unconstrained demand using ML models (deep learning, LGBM, …) trained on historical data in the form of time-series
• under-constraints optimizations solved using linear programming techniques

With the acceleration of our growth, the team is now entering a scaling phase where we will face the challenge to stay agile in terms of innovation while supporting and monitoring closely the in-production algorithms. To address this issue, we have organized the work around Kubeflow: a ML ops tool empowering data scientists to focus on high-added value tasks by maintaining a common framework and re-usable/ factorized components for all team members.

In this context, Wiremind is now looking for a head of ML capable of making necessary evolution on this existing framework and processes, developing and maintaining scalable ML pipelines and participating in integrating the pythie API into the real time architecture of the final application and growing the team as Wiremind expands.

WHAT YOU WILL DO

In a team shaped to have all profiles necessary to constitute an autonomous departement (devops, software eng., data eng., IA, Operational research), you will be responsible for :

• Leading the maintenance, development and research of new ML models for dataset that can span to billions of lines through reproducible, well documented and versioned pipelines
• Shaping the evolution of our general ML training framework and its components (A/B testing, validation, feature extraction, training…)
• Mentoring junior team member and onboard them in the framework
• Ensuring code quality, factorization and documentation
• Having a role in operations by developing and monitoring the link between Pythie and the final applications
• Participating in business analysis in par with our customer success team and the final clients

TECHNICAL STACK

• Python 3.7+
• KubeFlow over an auto-scaled kubernetes cluster for orchestration
• Druid as datastore
• Redash for business analysis
• Common ML librairies (tensorflow, lgbm, pandas, dash…)
• Gitlab for continuous delivery

WHAT IS IMPORTANT TO US

• Strong computer-science background in python with an interest for code quality and good practices (unittesting, pep8, typing)
• A first relevant experience in a data team with production ready pipelines using DAGs
• Knowledge about at least one deep learning framework: tensorflow or pytorch
• A pragmatic approach to ML where testing and frequent deliveries of incremental gains supported by validation / alerting processes to avoid regression is preferred to a long tunneled research process

WHAT WILL BE A PLUS

• A first experience with airflow or kubeflow
• A first experience modelling time series

The position is based in the city-center of Paris (métro Etienne Marcel) in our renovated office. We offer attractive pay packages (depending on your profile), closely linked to your performance and the company’s growth.