Back to Projects
T
Engineering

The Last Message

Full-Stack Developer

NestJSTypeORMPostgreSQLNext.jsReact NativePython

About the Project

The Last Message is a digital publishing platform built around a specific book — a web and mobile reader with paid access. The interesting half is the content pipeline: the source material arrived as a PDF and had to be extracted, split into sections, structured, and imported into the database before any of it could be served as navigable, searchable content.

What It's Made Of

Content ingestion pipeline

Python · PDF extraction · multi-pass

The interesting half of the project. The source material arrived as a PDF with inconsistent section markers and formatting that broke naive parsing, so extraction runs in several passes with manual verification steps rather than one script, segmenting content into sections before import. Kept as standalone scripts rather than folded into the API, because ingestion is a one-off operation and should not add dependencies to a service that just serves content.

API

NestJS · TypeORM · PostgreSQL

Serves the book as navigable, searchable content, with chapters, sections and messages modelled as independently addressable entities so a reader can deep-link and resume. It also owns purchase state and access control.

Web reader

Next.js

The reading experience: section navigation, progress tracking, and search across the text, with Paystack handling purchase and gating access to paid sections.

Mobile reader

React Native

The same reading experience on a phone, written against the same API as the web client so content structure, progress and entitlements stay consistent across both.

Key Highlights

  • Wrote the Python ingestion pipeline that extracts PDF content, segments it into sections, and imports it into PostgreSQL
  • Built the reading experience in Next.js with section navigation and progress tracking
  • Integrated Paystack for purchase and access control
  • Built the React Native mobile reader against the same API
  • Modelled the content schema in TypeORM so chapters, sections, and messages stay independently addressable

Technical Challenges

PDF extraction is deceptively hard — the source had inconsistent section markers and formatting that broke naive parsing, so the pipeline needed several passes and manual verification steps rather than a single script. I kept the pipeline as standalone Python scripts rather than folding it into the API, because ingestion is a one-off operation that shouldn't add dependencies to a service that just serves content.