Updating Your ML Serving Stack

Harshit Agarwal, Senior Machine Learning Engineer at Faire, discusses how Faire transitioned from a traditional infrastructure to a modern, flexible ML deployment and serving stack. The talk covers the setup of a system that supports various model types, ensuring operational excellence and scalability in e-commerce. Key points include infrastructure as code, CI/CD pipelines, automated testing, user-friendly model release management, and built-in observability and monitoring for model performance and reliability. This approach enables the team to quickly bring new ideas to life while maintaining operational stability.