What we do
Tools
Industries
Department
How we workLet's talk →
Home/Python Developer
Python Development

Hiring a Python Developer for the Script Your Business Now Runs On

Python is often the first thing someone reaches for to clean a spreadsheet or copy data between two systems. That's its strength. The trouble starts a year later, when the script builds the report the managers read every Monday and still only runs on one laptop, with one person who knows how to start it.

A script someone runs, or a system the business depends on?

Same language, very different job. Which side you're on decides what you're hiring for.

Still a script

Run by hand, when someone needs it

  • One person runs it, watches it finish and checks the output by eye.
  • If it doesn't run this week, someone notices and nothing important stops.
  • It reads a file and writes a file. No logins to other systems.
Now a system

Runs on a schedule, and other work waits for it

  • A report, an invoice run or a stock update depends on its output.
  • It logs into your CRM, accounting or store with stored credentials.
  • If it stops, the first sign is a number that looks wrong days later.

If even one point on the right is true, the code is the smaller part of the job. What it needs is a home: somewhere to run, a fixed environment, its own credentials, and a way to tell someone when it fails.

Moving a Python script off one person's laptop

None of this changes what the script does. It changes whether anyone else could run it tomorrow.

  1. 1

    Put the code in a repository

    Not a shared drive and not an email attachment. A repository your company owns, so every change has a history and nobody edits the live copy by accident.

  2. 2

    Write down the environment

    python -m venv .venv pip freeze > requirements.txt

    Record the Python version and the exact library versions it runs with, in a lock file or a pinned requirements file. Then it builds the same way on any machine.

    Watch for: a script still on Python 2, which stopped receiving fixes in January 2020.

  3. 3

    Take the passwords out of the file

    Move API keys and logins into environment variables or a secrets store, and give the script its own account in each system instead of a person's.

  4. 4

    Run it somewhere that stays on

    A small server, a container or a scheduled cloud function, triggered by a schedule rather than someone remembering. Paths like C:\Users\dave\Downloads become settings.

  5. 5

    Make failure loud

    Log each run, and send an alert to a shared channel when it fails or when it succeeds with suspiciously little data. A run that processed zero rows is often the real failure.

  6. 6

    Test the parts that touch money

    A few tests on the calculations, the date handling and anything that writes to accounting. Those are where a library upgrade changes results without an error.

Why it works on their machine and nowhere else

What we find most often when a Python script is handed over.

An unpinned library moved underneath it

How you'll know: The script ran for a year, then failed or changed its numbers after someone installed something else.

Fix: Pin versions, then upgrade on purpose with the tests running. pandas in particular has changed defaults between major versions.

The Python version is past its end of life

How you'll know: It runs on whatever Python the laptop came with, and nobody knows which.

Fix: A new Python ships every October and each gets about five years of fixes. Python 3.10 has now reached its end, so anything older needs moving to a supported version.

It reads a file someone saves by hand

How you'll know: The run fails whenever the export is renamed, saved as .xlsx instead of .csv, or left open in Excel.

Fix: Fetch the data from the system's API where there is one. Where there isn't, check the file's columns before processing and stop with a clear message if they've changed.

It's quietly a database

How you'll know: Results are kept in a growing pile of CSV files that the next run reads back in.

Fix: Move that state into a real database table. It's a small change that ends a whole class of duplicate and missing-row problems.

Questions people ask

Do you build data pipelines, or mostly general scripts and apps?

Both. Much of our Python work is the layer between two systems: pulling orders from a store into accounting, cleaning supplier files, building reports from several sources. We also build small internal web tools, often with Django, when a team needs screens on top of the data.

What happens if our existing Python tool was built by a freelancer who left?

We get it running somewhere other than their machine first, which shows what it depends on. Then we read through it, list what it logs into and what it writes, and give you a short report on its state before proposing changes.

How do you catch a pipeline failure before it causes real damage?

Every run logs what it did and how many rows it handled. A failed run, a missed schedule or a run with far fewer rows than usual sends an alert to a shared channel your team already watches, rather than one person's inbox.

What do I actually get from the free quote?

A plain read of what the script does, what it needs to run and what's fragile about it, then a price to give it a proper home, with monitoring included. If it's really two systems that need connecting, the quote covers that route as well.

Should a script like this be rewritten in something other than Python?

Rarely. Python is a good fit for data work and system glue. What usually needs changing is how it's run and looked after, not the language. Bigger services sometimes move to a separate backend with a proper API.

What makes a junior Python developer hireable for business work?

Code that runs on someone else's machine. Beyond the language basics, that means virtual environments and pinned dependencies, reading and writing CSV and Excel files, calling a web API, and SQL. Hand a project to a friend and see whether they can run it from your notes alone.

Let's talk

Tell us the one thing your team does manually that eats up time. We read every message and reply within a day.