Sole data scientist at a mid-sized video game studio. No DS manager, no peers — I owned the entire ML lifecycle from data collection to model deployment to stakeholder communication.
Graph Neural Networks at Scale
The main project was predicting private user behavior from public social graph data across 60M+ anonymous user profiles. The model outputs directly informed product strategy and marketing targeting decisions for company leadership.
Beyond the GNN, I built models for clustering, churn prediction, look-alike audience targeting, and promotional timing optimization.
Infrastructure
- Data Processing
- 1.6TB+ via Apache Airflow and Spark across 20+ containers
- Compute
- Hybrid Proxmox/GCP deployment
- Databases
- PostgreSQL (relational), Neo4j (graph)
- Monitoring
- Airflow DAGs, PostgreSQL health checks
- Scale
- 60M+ anonymous user profiles, billions of data points
Analytics Dashboard
Built a full-stack React dashboard that translated model outputs into something non-technical leadership could act on — player segmentation, churn analysis, promotional impact, competitor analysis across the company's titles and key competitors.