All projects
Client & Production Projects

Bookize An eBook conversion and reading SaaS platform, with AI-powered summarization and proofreading tools.

A SaaS platform covering the full stack — document conversion, an online reader, AI-powered features like summarization and proofreading, and Stripe-based subscriptions — alongside the content and SEO work that grew its organic search visibility.

Bookize online reader with Alice in Wonderland open at chapter one

Client / Project

Bookize

Industry

SaaS — document conversion & AI-assisted writing tools

My Contribution

  • Next.js frontend + i18n
  • Backend architecture + PostgreSQL
  • Stripe subscriptions & webhooks
  • AI integration (Ollama)
  • Keycloak authentication
  • File conversion pipeline (Aspose.Words + Hangfire)
  • Content strategy & SEO
Bookize online reader with Alice in Wonderland open at chapter one

The Brief

Bookize is a SaaS product for converting and working with eBooks and documents — from format conversion to an online reader and AI-powered tools like summarization and proofreading — built out end to end, with product direction owned alongside the engineering.

View live

Built with: Next.js, TypeScript, C#, ASP.NET Core, PostgreSQL, Stripe, Keycloak, Aspose.Words, Hangfire, Ollama

Problem

The product needed to grow from a narrower conversion tool into a fuller SaaS platform: a modern, internationalized frontend, real subscription billing instead of ad hoc access, AI-powered reading tools (summarization, proofreading) to differentiate from plain converters, and a way to run slow, resource-heavy document conversions without blocking the rest of the product.

Solution

Rebuilt the frontend on Next.js with multi-language support from the start, designed the backend architecture in C#, and moved persistence onto PostgreSQL. Built an online reader that opens EPUB, MOBI, PDF and other formats directly in the browser, with highlights and notes. Added Stripe subscriptions with self-built webhook handling, Keycloak for authentication, and AI-powered reading features — including a text summarizer and a proofreading tool — integrated with an internally hosted Ollama model over plain HTTP calls rather than an agent framework. Document conversion runs through Aspose.Words, with heavier conversion jobs offloaded to Hangfire background processing so they don't block the product.

Feature direction, content, and ongoing improvements weren't handed down as a fixed spec — they came out of continuous self-directed research into what to add, refine, or drop, effectively running product decisions as well as building them.

Key Features

Conversion across eBook and document formats — EPUB, MOBI, AZW3, PDF, Word — with the file processed in the background and ready to download.

Bookize EPUB to PDF converter with an EPUB file ready to convert

An online reader that opens EPUB, MOBI, PDF, AZW3, Word and other formats directly in the browser, with highlights and notes kept alongside the book.

An AI summarizer that turns a long chapter into a short, readable summary, with adjustable length and optional quiz generation.

Bookize AI summarizer showing a summary of Alice in Wonderland's first chapter

An AI proofreader that finds grammar, spelling, and punctuation issues and explains each fix.

Bookize AI proofreader listing grammar and spelling fixes for a passage of text

Technical Approach

  • Frontend: Next.js with i18n built in from the start.
  • Backend: C# service layer designed and evolved from scratch, backed by PostgreSQL (introduced partway through the project).
  • Reader: in-browser reading for EPUB, MOBI, PDF, AZW3, Word and other formats, with highlights and notes.
  • Auth: Keycloak.
  • Payments: Stripe subscriptions, with webhook handling implemented and maintained directly.
  • Document processing: Aspose.Words for the core conversion logic, with longer-running conversions handed off to Hangfire so the request path stays responsive.
  • AI: A text summarizer and a proofreading tool integrated via direct HTTP calls to a self-hosted Ollama model — no agent framework, by design, given the scope.
  • Delivery: Deployment and infrastructure were owned by a separate DevOps function; day-to-day involvement extended to the project's gitlab-ci.yml pipeline.

Challenges

Keeping heavy conversions off the request path

document conversion can be slow and resource-intensive; moving it into Hangfire background jobs kept the product responsive without rearchitecting the whole request flow.

Owning Stripe end-to-end

subscriptions and webhook handling were designed and implemented from scratch, including the edge cases around plan changes and renewal state that don't show up until real usage hits them.

Shipping AI features without an agent framework

integrating the summarizer and proofreading tool through direct HTTP calls to Ollama meant handling structure, reliability, and failure cases explicitly, rather than leaning on an existing framework's guardrails.

Deciding direction, not just implementing it

feature scope, content, and technical direction had to be researched and decided as part of the work, rather than arriving as fixed requirements to build against.

My Role

Worked across the full stack: the Next.js frontend and its i18n setup, the online reader, the backend architecture and its move to PostgreSQL, Stripe subscriptions and webhooks, Keycloak authentication, the Aspose.Words/Hangfire conversion pipeline, and the Ollama-based AI integration. Also wrote product content with AI assistance and drove the SEO and feature-research work described below. Beyond implementation, took on the product side as well — deciding what to build, improve, or cut, rather than working from a fixed backlog. Deployment infrastructure was handled by a separate team, with direct involvement limited to the CI pipeline.

Conclusion

Bookize grew from a narrower conversion tool into a full subscription SaaS product — with AI-powered reading tools like summarization and proofreading, reliable handling of slow document conversions in the background, and a steadily growing base of organic search traffic behind it. The result wasn't one big rewrite; it was a product that kept moving in the right direction, both technically and in the market.

Real products, practical engineering, and problems worth solving.

Selected work across .NET, full-stack SaaS, AI integrations, developer tooling, React, React Native, and more.

Get In Touch