Warning: This section is under construction

GCP

What I’ve touched...

• Vision API
• Cloud Run
• Artifact Registry
• Kubernetes Engine

Vision API

I have pretty limited experience with GCP. In my time at Capsule (around Winter 2020), I built a proof of concept for insurance card image text extraction. I used the Vision API in conjunction with Cloud Functions V1 to make a serverless API endpoint that would return JSON data of extracted text. I only ran minimal model training on a small dataset of images, and that was quite expensive and highly manual (I had to use the web UI and manually drag boxes around specific insurance card fields). I never got around to implementing the association of text to insurance fields, like "bin number" and "group number", and the project was eventually derailed by internal misalignment — basically it was cut short, but was certainly fun while it lasted.

Kubernetes Engine

I’m currently running a Waypoint server and runner on an Autopilot Kubernetes cluster, hosted on Kubernetes Engine. It costs about $30/month.

Cloud Run

I only recently (July 2022) came across Cloud Run. It’s basically Docker containers on Lambda. Here's a Deno fresh instance running on Cloud Run:

• https://deno-fresh-r6og3wfz2a-uc.a.run.app

Misc

GCP x Waypoint

https://myshittycode.com/2019/10/30/gcp-terraform-google-could-not-find-default-credentials-error/