At appflame, I built a feature-based model and classifier for email templates that feeds rotation outcomes back into ranking. It enabled cold-start ranking for templates with no performance history and surfaced patterns linked to stronger or weaker results.
I translated product insights into A/B tests of email-sending rules and analyzed more than 20 experiments. The best-performing experiment increased mail revenue per user by 200%+, while other tests improved Click Rate or reduced unsubscribe rate without a negative revenue impact.
I also developed a root cause analysis system for conversion metrics and built automated template rotation to respond faster to performance drops. I contributed to an AI-friendly BI system, an analytics agent’s semantic layer, experimentation methods including CUPED, and data marts and Prefect workflows for monitoring and alerting.
