GopherCon 2025: Supercharging ML Pipelines with Go - Vaidehi Thete

Why watch

This GopherCon talk demonstrates how Go's concurrency model, shared memory IPC, and efficient queuing reduced ML inference times from hours to 10-15 minutes. It covers architecture decoupling operational responsibilities (feature retrieval, queuing, shared memory) from ML ranking tasks, with real-world benchmarks and implementation details. Designed for software engineers and system architects building high-performance ML inference systems.