The company is data-driven and uses data and algorithms across its games and its marketing.
The team sits in the Data Office's Player Domain and works on player safety and security. It builds and deploys machine learning systems, such as graph neural networks that detect sanction circumvention and transformer models that flag toxic chat in real time.
The work covers NLP, anomaly detection, graph learning and large language models. The company is a global games maker with teams around the world.
Responsibilities
Build scalable machine learning systems that keep players safe, taking projects from research through to production at scale.
Design, train and deploy models for player safety and security across NLP, anomaly detection and graph-based learning.
Turn the latest research in deep learning, NLP and graph learning into production-ready methods, coding them from scratch when needed.
Own your projects end to end: work with stakeholders, set success metrics, monitor live systems and keep iterating.
Write, tune and ship high-quality code that runs at scale, using MLOps practices such as CI/CD, model versioning and observability.
Work with data and software engineers to move models and prediction pipelines into production on internal or public cloud infrastructure.
Requirements
You have a track record of taking machine learning models to production, and you move comfortably between research and engineering.
Strong Python skills: clean code, testing, documentation and CI/CD.
A track record of putting ML models into production and watching them afterwards, covering model serving and drift monitoring.
Hands-on work with NLP or LLMs, such as fine-tuning transformers, prompt engineering, RAG or LLM pipelines.
Experience with graph ML (GNNs) is a strong plus.
Knowledge of anomaly detection or fraud and trust and safety systems in production.
Experience with distributed data processing in Spark and with cloud data platforms. Familiarity with Databricks (DAB, Delta Live Tables) or Azure services is a plus.
Strong communication skills. English is mandatory for presentations, cross-team work and reporting.
Ideal candidate: experience with token classification (NER, sequence labelling) and/or hands-on GNN training (node classification, link prediction, graph construction pipelines).
Conditions
Office based in Saint Mandé, reachable on metro line 1 at Saint Mandé station.
Equal opportunity: hiring is open to everyone regardless of gender, ethnicity, religion, sexual orientation, social status, disability or age.
Benefits
Profit sharing and a yearly company savings plan.
25 days of paid time off plus 12 additional paid days off.
The company pays 50% of your transport pass.
Lunch vouchers at 9€ per day.
Healthcare covering you and your family.
Maternity leave of 20 weeks and paternity or co-parental leave of 7 weeks.