LiveMLP - ML Platform for Assisting Contact Center Agents in Real-Time

Explore a conference talk that delves into the development of a real-time Machine Learning platform designed to assist contact center agents during live calls. Learn about the challenges faced by contact centers, including high call volumes, agent stress, and attrition rates. Discover how traditional performance improvement methods fall short and why a real-time system is necessary. Examine the engineering challenges involved in creating such a system, including throughput vs latency trade-offs and fault tolerance. Gain insights into the differences between real-time ML systems and post-call systems, particularly in terms of batch vs non-batched inference and context. Follow the speaker's journey in developing a horizontally scalable, low-latency ML platform, and understand the approaches taken to ensure robustness and stability. Witness the platform's efficacy through implementation and load testing results with up to 10,000 concurrent calls. Conclude by learning how this innovative system positively impacts business metrics in real-world scenarios.

SREcon23 Asia/Pacific - LiveMLP: ML Platform for Assisting Contact Center Agents in Real-Time