A multi-week reporting slog, down to a few minutes

python airflow polars

Cebir builds assessment tests. Every cycle, they compute statistics on test takings and turn them into detailed reports — work that used to be done by hand and often took several weeks to finish.

The process was split across sub-processes, with a lot of copy-pasting of both code and data from one step to feed the next. Every handoff was a chance for a mistake, and the whole thing was slow.

What shipped

An Airflow-orchestrated pipeline that does the heavy data work in Polars and generates the finished reports with python-docx and openpyxl. The manual handoffs between steps are gone — the chain runs end to end on its own.

Constraint that shaped it
It had to slot into how the team already worked and stay maintainable by them afterwards — so the right tool for each step, and no cleverness that would need me around to keep it running.

Result

A process that took weeks now runs in minutes, and the manual copy-pasting that caused most of the errors is gone entirely.

At a glance
clientCebir — Kortenberg year2024 – 25 duration~8 months rolesolo — freelance
runtime
minutes
from several weeks
manual steps
0
Manual effort per cycle
working days · before vs after
0 6 12 before after
working days / cycle (approx)
Stack
python airflow polars python-docx openpyxl
want something similar?
Tell me what's slow, manual, or fragile
Start a project