We can't find the internet
Attempting to reconnect
Something went wrong!
Hang in there while we get back on track
Petros Syntelis - Modelling of agent-customer pairing outcomes to optimise call centre performance
Learn how Petros Syntelis applies causal inference and machine learning to optimize call centre performance by modelling agent-customer pairing outcomes, ensuring data integrity and safety, and predicting the best pairing for specific KPIs.
- Modelling of agent-customer pairing outcomes requires causal inference techniques to identify the cause of an effect.
- The goal is to identify the best pairing of customer and agent to optimize a specific KPI.
- Causal modelling is used to estimate the uplift, which is the expected impact of an intervention.
- The IVR (Interactive Voice Response) is a critical component of the call centre system, requiring abstraction and decoupling from business logic.
- To move to a productionized system, Pydantic is used to ensure data integrity and safety.
- The agent updater is a system that makes bulk changes, requiring careful testing and measurement of latency.
- The system uses Apache Airflow for data pipelines and dbt for data transformation.
- To optimize the system, randomized experiments are used to minimize the impact of confounders.
- The Canary is an application that runs in GCP, detecting changes and propagating alerts to the development team.
- The system uses a combination of machine learning and causal modelling to predict the outcome.
- The modeling process involves estimating the uplift and identifying the most important features.
- The system uses a data-driven approach to optimize the pairing of customers and agents.
- The engineering team worked closely with data scientists to develop the system.
- The system was deployed in a production environment and continuously monitored for errors and latency.