Reasoning at Scale: Knowledge Graphs, Hybrid Search, MCTS, and Kubernetes

This talk walks through the design and implementation of Bird Watchers’ Emporium, a production-grade reasoning system built on knowledge graphs, hybrid search, and MCTS-based planning. I’ll cover evaluation harness design, fine-tuning workflows, and system orchestration using Kubernetes and Kafka. The talk briefly touches on the AWS/Terraform/React stack, with emphasis on the components that enabled scalable reasoning and reliable deployment.

About the Speaker
Liam Barstad is a Machine Learning Engineer with 8 years of experience spanning web development, data engineering, devops and Machine Learning. He currently specializes in scaling and tuning LLM-based applications.