How I Automated The Worst Part of My Job (Until it Tried To Eat Me Alive)

Every software team has that one recurring task that nobody wants to do - but everybody depends on. For us, it was exporting reports from Elasticsearch. Simple enough at first: take a saved search, export the results to CSV, and email it on a schedule. But things got complicated. Business logic crept in. Filters got conditional. Fields needed transformation. Suddenly, Elasticsearch’s out-of-the-box tools weren’t cutting it and users still expected their reports to land on time.

This talk is about how I used Python to automate the worst part of my job: building and maintaining a custom reporting system on top of Elasticsearch. I’ll walk through the journey of going from a quick-and-dirty script to a modular, cron-based system - and how one particularly complex request pushed it beyond its limits. Rather than refactor endlessly, I took a different approach: I cloned the system and built a purpose-specific version just for that report. It wasn’t elegant, but it worked - it taught me a lot about pragmatism, system boundaries, and what “good enough” looks like in production.

This talk is for backend developers, DevOps engineers, and anyone who’s had to glue tools together when official solutions fall short. If you’ve ever worked with, or plan to work with Elasticsearch, cron jobs or Python-powered automation, you’ll walk away with practical ideas, architectural insights, and hopefully a few laughs.