At Comviva Technologies, I work on-site with Ooredoo Tunisia’s marketing and CVM teams, taking ML campaigns from requirements and KPI definition through delivery and impact measurement. I delivered six production models, and ML-driven target groups have consistently outperformed the Universal Control Group.
I retrained the Next Best Offer model and combined it with Next Best Time targeting, scaling NBO to one million subscribers and achieving 18% higher offer take-up than the Universal Control Group. I also built a Winback propensity model that cut discount spend by 15% while improving the winback rate by 9%.
My work includes recharge validity, digital and OTT propensity, offer recommendation, and RFMI-based customer segmentation. Earlier at Comviva Technologies, I automated Python and SQL pipelines, reducing manual effort by 30%; at Tatwa Technologies, I built a production RAG system for enterprise query resolution. I’ve also built personal projects in agentic AI and conversational RAG.
