Afiniti develops AI-driven systems that optimize human interactions, specifically by enhancing the way customers are matched with agents in real-time contact centers. During my tenure, I led core machine learning efforts focused on driving measurable business impact through data-driven behavioral pairing.

Contact centers handle millions of interactions daily, but traditional call routing doesn't account for the compatibility between customers and agents. Afiniti's behavioral pairing technology aims to improve outcomes by intelligently matching individuals based on predicted interaction success — increasing conversion rates, customer satisfaction, and operational efficiency.

Tracking

• Built and refined supervised learning models that predict key outcomes (e.g. conversion, retention, NPS) based on behavioral, demographic, and historical interaction data.
• Used gradient boosting (XGBoost, LightGBM), logistic regression, and survival analysis techniques to forecast call outcomes.
• Led feature engineering efforts using customer CRM data, interaction logs, and third-party enrichment to capture latent behavioral signals.

 

ML Deployment & Real-Time Inference

• Collaborated with engineering teams to productionize models with low-latency constraints, integrating them into live call-routing systems.
• Implemented model versioning, real-time scoring APIs, and drift detection mechanisms to ensure long-term performance stability.
• Used Docker and Kubernetes for containerized deployment on scalable infrastructure.

 

Experimentation & Business Metrics

• Designed and monitored controlled experiments (A/B testing) to evaluate the business impact of behavioral pairing strategies.
• Quantified improvements in conversion rates, average handling time (AHT), call resolution, and customer retention metrics.
• Worked closely with client-facing teams to tailor models to specific industries, such as:
• Telecom (e.g. NOS): focused on churn prediction and upgrade conversion.
• Insurance (e.g. Liberty Mutual): optimized retention and cross-sell conversations.

 

Global Collaboration & Custom Solutions

• Partnered with teams across Europe, APAC, and the Americas to customize ML pipelines for regional compliance, data availability, and cultural variation.
• Handled multi-lingual NLP challenges when applicable (e.g., intent extraction or sentiment analysis from agent transcripts).

 

Data Analysis & Visualization

• Built interactive dashboards (using tools like Tableau and Plotly Dash) to visualize model insights, customer-agent dynamics, and experiment results for stakeholders.
• Delivered data storytelling presentations to product, strategy, and executive teams to support data-driven decision-making.