A hands-on deep-dive into production-grade GPU model serving — deploying a handwriting-recognition (MNIST) model behind an HTTP API on two clouds at once, consumed by a small desktop client that lets you draw a digit and pick which backend answers it.

Why two clouds: built the same serving pattern on AWS SageMaker and Alibaba PAI-EAS, both fronting Triton Inference Server, with all infrastructure defined as code (AWS CDK and Alibaba ROS CDK).

A Git-based promotion strategy for a lead-time forecasting model across separate dev, integration, and production AWS accounts — solving training-data curation, pipeline orchestration, and validated-model promotion as one connected system.

Design choice: SageMaker Pipelines over Step Functions for native MLOps support — pipeline caching, model lineage, and registry integration out of the box. Models carry champion/challenger aliases and are promoted explicitly, never silently overwritten.

Connecting a conversational front-door (Open WebUI) to durable, centralized storage for departmental survey data — replacing hours of manual, repeated data collection with a chatbot backed by a serverless AWS API.

Impact: saves one full-time employee, per the requesting department's own estimate. Validated with a two-day low-code prototype before committing to the production build; presented to skip-level leadership.