Building the backend — Node.js and Python FastAPI

A polished frontend is only half a real product. The moment a project needs to store data, talk to other services, process payments or run logic you can’t trust to the browser, it needs a backend. I build those with Node.js and with Python’s FastAPI — and choosing between them is a big part of getting it right.

My two go-to backends

Node.js (with Express or Fastify) — my default when the backend lives close to a web app:

• Same language as the frontend (TypeScript), so I share types and validation between client and server.
• Great for API routes, webhooks (Midtrans, Stripe), and talking to the same services the frontend uses.
• Pairs naturally with Next.js API routes and serverless functions.

Python + FastAPI — my go-to when the backend is data- or AI-heavy:

• Automatic, typed request/response validation with Pydantic — fewer bugs, less boilerplate.
• Auto-generated OpenAPI/Swagger docs out of the box, so the API is documented and testable from day one.
• Async support for fast, concurrent endpoints, and direct access to Python’s data and ML ecosystem.

A typical FastAPI endpoint

The thing I like most about FastAPI is how little ceremony it takes to get a typed, validated, documented endpoint:

from fastapi import FastAPI
from pydantic import BaseModel

app = FastAPI()

class Booking(BaseModel):
    name: str
    tour: str
    guests: int

@app.post("/bookings")
async def create_booking(booking: Booking):
    # validated automatically; persist to the database here
    return {"status": "received", "tour": booking.tour}

That’s a fully validated POST endpoint with interactive docs at /docs — no extra wiring.

How I choose

• Close to the web app, sharing types, handling webhooks? → Node.js.
• Data processing, analytics, AI, or a service other teams will call? → FastAPI.

Either way, the frontend stays fast and the heavy lifting happens in a backend that’s built for the job. That’s what full-stack actually means in practice: not one favourite tool, but the right one for each part of the system.