Thursday, August 22

3:30 pm - 4:10 pm

Deploy a Serverless Language Model API

Inline Stage


About the event

This session will serve as an introduction to using GCP's Cloud Run as a service to deploy and host a serverless machine learning prediction API. Python’s FastAI libraries will be used for natural language processing predictions which will be easily triggered using HTTPS requests. During a live demonstration, Docker images will be built to contain multiple exported models. Then the serverless application will be deployed based on this image. Prediction requests will then be made using Postman. Additional topics will include transfer learning and alternative natural language services. By seeing the workflow first hand, attendees will be able to build upon this example to deploy their own serverless applications and cost-effectively host custom trained machine learning models.

Speakers

James Kennedy

James is a new Pittsburgher and recently joined Mitsubishi Electric to design and sell battery systems for the power grid. He previously worked for several years as an evaluation engineer and data scientist determining the impacts and performance of energy resources on behalf of utilities, public commissions, and state agencies.