Vertex AI MLOps: Model Evaluation Framework

Project Overview

Established a cloud-native MLOps framework on Google Cloud's Vertex AI for responsible deployment of fine-tuned generative models. Built orchestration with Vertex AI Pipelines and Cloud Build, maintained artifacts in Cloud Storage, and exported metrics to BigQuery. Implemented custom monitoring tracking content-aware metrics (latency P95, refusal rates, citation faithfulness) with automated evaluation via Cloud Scheduler.

Technology Stack