Kaustubh.dev

Wallxy · work · 2024 – present

Async Worker Pipelines & Platform Ops

The background infrastructure behind Wallxy: Pub/Sub workers on Cloud Run, Dockerized workers on a Linux VPS, cron-built feeds, and the monitoring around them.

Visit Wallxy
Role
Owned all infrastructure & deployments
Timeline
2024 – present
Status
In production
GCP Pub/SubCloud RunDockerLinux VPSRedisMongoDBVercelSentryPostHogCloudConvert

I own Wallxy's infrastructure end to end: production deployments on Vercel, a Linux VPS running Dockerized backend workers, Google Cloud (APIs, Cloud Run, Pub/Sub), and application health through Sentry and PostHog.

worker pipelines on Pub/Sub + Cloud Run
5
content types in the Explore feed
5
developers whose deploys and merges I manage
3
runtimes: Vercel · Cloud Run · VPS
3

The problem

A design-collaboration SaaS does a lot of work nobody should wait for: running campaigns, converting PDFs into DWG files, expiring subscriptions, sending email, and billing for usage. It also needs a fast, fresh discovery feed.

Architecture

A simplified view of the async side of the platform.

App

  1. Next.js on Vercel

    UI + API routes

  2. MongoDB

    primary data

Queue & workers

  1. GCP Pub/Sub

    topics per job type

  2. Cloud Run workers

    campaigns · PDF→DWG · expiry · email · usage billing

Feeds & ops

  1. Docker on Linux VPS

    backend & cron feed workers

  2. Redis

    Explore feed cache

  3. Sentry + PostHog

    errors & product analytics

Worker pipelines

  • Campaign execution
  • PDF-to-DWG conversion
  • Subscription expiry
  • Email delivery
  • Usage-based billing

Explore feed

A multi-type feed covering ideas, catalog, gallery, brands and people, with Redis caching, real-time search, and trending content computed by cron-based feed workers.

Operating it

I manage production deployments on Vercel and the VPS, administer Google Cloud Console, and watch application health in Sentry and PostHog. As tech lead I also run sprint planning and code reviews, and handle every Git merge and branch for a team of three.

Key decisions

01

Keep slow work off the request path

Anything that can take seconds or minutes goes onto a Pub/Sub topic and a worker, so the UI stays fast.

02

Right host for each job

Serverless Cloud Run for bursty queue work, a Dockerized VPS for long-running and cron workers, and Vercel for the Next.js app.

03

Cache the feed, compute trends offline

Redis serves the Explore feed quickly, while cron workers precompute trending content instead of calculating it per request.

Want something like this built, or someone who builds it?

Tell me about your project, or grab my résumé if you're hiring. The code is private; I'm happy to walk through it in an interview.

Next case studyAI Social Agent