Stavros Niafas
Stavros Niafas is a ML engineer and an MLflow ambassador. As a ML practitioner has demonstrated experience in both R&D and production settings, from driving ML experiments and PoC's into mature codebases to support e2e ML pipelines and systems. Stavros is also actively engaged in FOSS communities and conferences, MLOps and systems engineering.
Session
Machine learning models require frequent updates to maintain accuracy and consistency. Deploying model updates efficiently, without service interruption presents a significant challenge. This talk demonstrates a Pythonic approach to dynamically update MLflow models in production using AWS Lambda. We'll explore how to leverage AWS Lambda serverless architecture, along with Python and Boto3, to "hot-swap" models in the wild. The presentation will cover packaging models with MLflow, deploying them to Lambda and implementing a dynamic update mechanism through efficient model persistence.