qv* / project

Reflect ML Platform

“A production ML platform for a series A startup.”

2025 / for reflect / in-progress

the problem

Reflect was running ML inference on a hand-rolled stack that broke every other week.

the solution

I rebuilt it on a modern serving stack with autoscaling, model versioning, and a clean handoff between research and production.

human contribution

Architected the platform. AI handled the boilerplate around the deploy pipeline.

built with

cursorpythonpytorchkubernetes

Hire Vikram for something like this.

partially-available · $200/h · python