Amrutha Gujjar

Waldium hosts thousands of blogs on custom domains, all from a single Next.js deployment. Vercel's edge network, programmatic domain API, and automatic SSL provisioning make this seamless. Every tenant gets instant global performance without us managing servers. Their deployment API lets us provision domains programmatically. SSL certificates just... work. We're basically building Vercel for content. Fitting that we're building on Vercel. If you're building multi-tenant SaaS, their infrastructure is unmatched.

Waldium is multi-tenant, thousands of blogs and posts posts, all in one Postgres database. Neon's serverless Postgres scales instantly without us thinking about it. We like the Database branching. Every time we test a migration or new feature, we branch production data and test safely. No more "hope this migration works in prod." Connection pooling is built-in. Cold starts are sub-100ms. The free tier is absurdly generous for early stage. If you're building on Postgres and haven't tried Neon, you're missing out

Shoutout to @umami_software 📊 Every Waldium blog gets analytics. But we didn't want to force Google Analytics on everyone Umami is self-hosted, open source, privacy-first analytics. No cookies. No personal data collection. GDPR/CCPA compliant by default. And it's *fast*—2KB script vs GA's 45KB. We provision a unique website ID per tenant, track everything server-side, and give users clean analytics without ethical compromise. If you care about privacy but need analytics, Umami is the answer. Underrated in the indie hacker community.

It brings the ergonomics of DuckDB to the cloud. We use DuckDB locally in Pyodide for browser-based apps, and MotherDuck lets users transition easily to larger, cloud-scale datasets without rewriting queries. Same SQL, better scalability. If you’re already using DuckDB, integrating MotherDuck is almost zero effort and unlocks team sharing, auth, and query history. We also liked how easy it was to prototype with their Python SDK.

Running full Python in the browser means users can create and run apps without any backend. It’s magical. Preswald uses Pyodide to support live preview, AI refactors, and stateful execution, all offline. We considered Brython, Transcrypt, running a server and syncing via WebSockets. Startup time and memory can be a pain. lazy-load what you can, and show a clear loading UI. Also: wrap filesystem access in a clean API. Pyodide's internal FS can be tricky.

It’s insanely fast for local analytical workloads, and integrates beautifully with Pandas. Perfect for CSVs and Parquet files in browser-based workflows. It also doesn’t need a server or setup which is a huge win for interactive demos. DuckDB + Pyodide works surprisingly well. You can compile DuckDB to WebAssembly and run SQL queries in-browser with zero backend.

We pull high-quality images from the Unsplash API to seed our demo environments. This gives users instant visual context without any manual setup. Tip: It’s free and has a generous quota. We explored Pexels and Getty, but Unsplash had the best dev tools and licensing model for early-stage use.

Vercel handles our frontend deploys with zero-config CI/CD. Their edge functions keep load times fast and let us run server logic close to the user. Tip: Great DX for Next.js apps. We looked at Netlify and Cloudflare — Vercel won on speed and how tightly it integrates with our frontend stack

We use Supabase Storage and Postgres to manage all user-generated worlds and metadata. The built-in auth and instant APIs let us ship fast without standing up custom backend services. Tip: Works best if your data model fits cleanly into Postgres. We considered Firebase, but Supabase gave us SQL and easier self-hosting if we need it.