FastAPI services
Typed request and response models with Pydantic, automatic OpenAPI docs, async endpoints where I/O justifies them, and dependency injection for auth and database sessions.
Web & Application Development
Python development
Python covers three distinct kinds of work for me: web services with FastAPI or Django, data work with pandas and friends, and the automation glue that runs on a server and stops somebody doing a task by hand every week.
The third category is undervalued. A sixty-line script that reconciles two exports, or watches a directory and pushes files to an API, often returns more than a feature would, and it is cheap to build. I mention it because clients rarely think to ask.
How I work on it
Process
Adjusted to the job, but this is the shape it usually takes.
-
Pin the environment
A lockfile and a defined Python version from the start. "Works on my machine" is almost always a dependency resolution difference.
-
Types where they pay
Type hints on public interfaces and data models, checked in CI. Not dogmatic coverage of every local variable.
-
Test the transformations
Data logic gets unit tests with real awkward sample rows. The empty string, the wrong date format, the duplicate ID.
-
Run it somewhere real
Deployed under supervision with logging, not left in a terminal session that dies with the SSH connection.
Questions
Things people ask
FastAPI when the deliverable is an API and the consumers are other services or a JavaScript frontend. Django when you want the admin, the ORM, sessions, templates and auth to already exist. That back office is worth a lot on internal tools.
cPanel has a "Setup Python App" feature built on Passenger, and simple WSGI apps do run there. It is workable for low-traffic tools. Anything with Celery workers or heavy dependencies belongs on a VPS.
I do applied work, using existing models through APIs, embeddings and retrieval, classification with off-the-shelf libraries. Training novel models from scratch is a research specialism and I would point you elsewhere for that.