ML Scientist

At Metica, we let game studios focus on making great games – while we take care of growing them. We're an AI-native growth partner for mobile games studios, unifying user acquisition and in-game monetisation so games grow without studios having to build that expertise in-house.

We're built by founders and an early team on their third venture together (previously built and exited to King and Apple), backed by Play Ventures, 13books Capital and Firstminute Capital. We're already working with some of the biggest names in mobile gaming, and are entering our scaling phase – which means real ownership for whoever joins us next.

Meet Your Hiring Manager

Hiring manager: Puli Liyanagama, CTO

The Role

You'll own ML systems end to end – from hypothesis through experimentation, training, evaluation, deployment, monitoring and iteration. You'll write production-quality training and inference code, own your releases, and be accountable for the business metrics your models improve.

You Will

• Own ML systems end to end – hypothesis, experimentation, training, evaluation, deployment, monitoring and iteration.

• Write production-quality training and inference code, own releases, and be accountable for the business metrics your models improve.

• Design, execute and interpret A/B and multivariate experiments – power analysis, guardrails, exposure/assignment strategy and statistical significance.

• Apply contextual bandits, causal inference, uplift modelling, off-policy evaluation and propensity scoring to optimise real-world decisions.

• Analyse large-scale datasets, diagnose changes in model or product performance, and turn findings into clear recommendations for engineering, product and commercial teams.

You Need

• Expert Python and SQL, with experience building production ML systems and data pipelines (PyTorch, XGBoost, Spark/PySpark, Apache Iceberg, Ray, MLflow, SageMaker, Airflow, or equivalent).

• Deep experimentation and statistical inference skills, and a strong grasp of optimisation under uncertainty (bandits, causal inference, off-policy evaluation).

• Data-driven and self-driven – comfortable with ambiguity and able to work across engineering, product and customer-facing teams.

• Demonstrated ownership of production business metrics, not just research models, is a plus – as is effective use of AI-assisted and agentic software engineering workflows.