PREZA, School Presentation Sites with 80,000+ Generated PDFs

Tech stack

WordPressS3SelectelPDFScrapingAutomation

Links

Services

→Grabbers and parsers
→Automated multi-threaded presentation generation
→PDF production and storage
→SEO-friendly site structure

Deliverables

✓More than 80,000 generated PDF presentations
✓Storage of more than 300 GB in Selectel cold S3
✓Two sites: preza.kz and preza.top

Challenge

A project of sites with ready-made school presentations: preza.kz for Kazakhstan and preza.top in Russian. Each covers every subject and every grade, from primary to high school, in .pptx and .pdf. To be useful, the sites needed tens of thousands of presentations, a place to keep them all, and search engines able to find them.

Grabbers, Parsers and Generation

The first task was to generate a huge number of presentations in Kazakh. Later the same approach was applied to Russian.

Create presentation dialog with a topic in Kazakh, Mood in music — A presentation topic in Kazakh

I wrote the grabbers and parsers that fed the generation, and the pipeline that turned it all into files: in total more than 80,000 PDF documents, each saved on the server and opened to schoolchildren on the sites. For generation we used different solutions, ready-made platforms among them, and automated the work with them so presentations were created in many threads at once without anyone clicking through each one.

A presentation generator building a Kazakh-language slide deck step by step: slides, text, images — Automated generation on one of the platforms

The slides followed a set template: a few typical slide types, such as a cover, an introduction, text, text with a photo, a chart and a table, laid out in a programmed order. To make the text the neural network generated fit that structure, each template had its own dedicated prompt, and for each one a set of images was generated to match the structure exactly.

Presentation plan in a generation service: ten slides, each tagged with its template type such as cover, introduction, text, text with photo, chart and table — Each slide is tied to a template type

Storage

At one stage of the work the volume passed 300 GB, too much to keep on the site's own server. The files moved to Selectel's cold S3 storage, a Russian cloud provider, where each presentation sits in its own folder as a .pptx, a .pdf and a text file.

Selectel S3 bucket for preza.kz: a presentation folder with its .pdf, .pptx and .txt files — A presentation in Selectel cold storage

Results on Video

In this video I show the results of the automatic presentation generation: the overall volume, what the presentations look like, which topics they cover, and more.

The Sites

Both sites are built so that search engines index them well: every presentation has its own page inside its subject, and the catalogue links subjects and grades together.

‹›

aA🔒preza.kz/

⇧▢+

Static snapshot — scrollable Safari preview

‹›

aA🔒preza.top/

⇧▢+