Machine Learning Engineer (Risk & Fraud)

At Eneba, we’re building an open, safe and sustainable marketplace for the gamers of today and tomorrow. Our marketplace supports close to 20m+ active users (and growing fast!), provides a level of trust, safety and market accessibility unparalleled to none. We’re proud of what we’ve accomplished in such a short time and look forward to sharing this journey with you. Join us as we continue to scale, diversify our portfolio, and grow with the evolving community of gamers.

About your team

We're a Data team, bringing together specialists passionate about turning data into reliable, actionable decisions across Eneba. We build the foundations—data pipelines, datasets, and platform capabilities—that make analytics and machine learning scalable and trustworthy for the whole organization.

In this role, you'll join our Risk-focused efforts and help evolve our fraud detection capabilities. The team is moving from a legacy, manual approach toward an end-to-end ML lifecycle with robust training, deployment, monitoring, and evaluation. You'll work on real-time decisioning use cases, partner closely with Risk and engineering stakeholders, and tackle the practical challenges of fraud modeling—imbalanced data, delayed/partial labels, and building strong feedback loops to continuously improve model performance.

Responsibilities

• Build and iterate on real-time fraud/risk models (e.g., gradient boosting and anomaly detection approaches) to score transactions during checkout and support Risk decisioning.
• Own the full ML lifecycle for fraud detection models: data exploration, feature engineering, training, evaluation, deployment, monitoring, and continuous improvement.
• Design robust evaluation strategies for rare-event, highly imbalanced data, including handling delayed/partial ground truth and defining metrics aligned with business constraints.
• Partner closely with Risk, backend, and Data/Platform teams to productionize models behind an API, integrate with the risk engine, and improve model-driven decision flows (pre-/post-authorization).