Node.js
One language from database to pixel — TypeScript on the server, sharing types with every client we ship.
What it is, and why we use it.
Node.js runs JavaScript on the server. Its superpower in a product studio isn't speed — it's that the API, the web app, and the mobile app's logic can share one language, one type system, and often the same validation code. Fewer translation layers, fewer bugs.
Node (with TypeScript, usually Fastify or Next.js API routes) is our default backend for product APIs. We reach for Go when raw throughput or tiny memory footprints matter, and Python when the workload is ML-shaped.
Node.js vs Python vs Go.
Node against Python and Go — the three backends we actually deploy, each with a clear lane.
| Dimension | Node.js | Python | Go |
|---|---|---|---|
| Language | TypeScript / JavaScript | Python | Go |
| Sweet spot | Product APIs, realtime, BFFs for apps | ML, data pipelines, scripting, AI services | High-throughput services, infrastructure |
| Concurrency model | Event loop — superb for I/O-heavy APIs | Asyncio capable; the GIL limits CPU-bound work | Goroutines — best-in-class, trivial to use |
| Type safety | Strong via TypeScript, shared with frontend | Gradual (type hints), often unenforced | Static, enforced, simple |
| Raw performance | Good for I/O; weak for CPU-bound work | Slowest of the three | Fastest of the three |
| Team leverage | Same language as every client we build — huge | The entire AI ecosystem speaks it | Small, deliberate language — easy to maintain |
Node wins when
- The backend mostly orchestrates I/O: databases, queues, third-party APIs
- Types shared end-to-end with React/React Native clients
- Realtime features — websockets and server-sent events are native territory
Python wins when
- The workload is ML inference, embeddings, or data science
- You're gluing together the AI ecosystem (it all speaks Python first)
- Notebook-to-production research workflows
Go wins when
- Tens of thousands of requests per second on small machines
- Long-running infrastructure: gateways, workers, streaming
- Deployment simplicity — one static binary, no runtime
Node is our default because it collapses the team: the engineer who built the screen can fix the endpoint. We split out Go services when the load profile demands it and Python services when the work is AI-shaped — all behind the same Postgres.
Thinking about Node.js?
Tell us what you're building — we'll tell you honestly whether Node.js is the right tool for it.