Vertex AI

Technologies

aiplatform Google Python SDK
Codefresh Pipelines
GitHub Actions
FastAPI
Docker

Summary

Evaluated and prototyped Vertex AI as a platform for serving machine learning models in a production environment. Focused on automated model deployment, online inference performance, and infrastructure scalability. The work led to key architectural decisions that significantly reduced projected infrastructure costs.

What did I do?

Assessed and de-risked ML inference platforms by evaluating Vertex AI for production use.
Identified key limitations for large-scale inference.
Authored a report that led the company to pivot to Anyscale before full deployment, avoiding costly misalignment and saving ~$40K per month in infrastructure costs.