AI / EdTech

AI-Powered Course Summary Generator

Hours of manual writing reduced to minutes — for a Dutch EdTech platform serving universities.

Live in production
Client
Dutch EdTech platform
Industry
EdTech / University study aids
Engagement
Full-stack product build, ongoing iteration
Timeline
2-week initial build, continuous refinement

The Challenge

The Dutch EdTech platform serves university students across the Netherlands who need condensed, readable summaries of dense course material to prepare for exams. Their internal team was manually creating these summaries — a slow, expensive process that did not scale across hundreds of courses and multiple universities.

They needed an internal tool their employees could use to generate high-quality, exam-ready summary documents on demand, in both Dutch and English, formatted as polished Word documents ready for student delivery.

What We Built

A web-based generator where the operator inputs the course name, study program, university, and pasted course information from the university website, then specifies a target page length between 30 and 200 pages. The system uses GPT to plan a chapter structure, generate chapter-by-chapter content in parallel, and export the result as a formatted Word document with inline images, equations, and proper chapter numbering.

Key engineering decisions

  • Parallel API orchestration

    Rather than generating chapters sequentially, we run API calls in parallel with a shared context layer, cutting generation time significantly for long documents.

  • Real-time LaTeX-to-image conversion

    Mathematical and statistical courses (epidemiology, biostatistics, probability) require properly rendered equations; we detect LaTeX in the generated content and convert it to images on the fly so the Word output is print-ready.

  • Single-input length control

    We replaced an early multi-field structure spec (min/max chapters, subchapters, page lengths) with a single "total pages" input, letting the AI optimize chapter depth and breakdown automatically. Cleaner UX, better outputs.

  • Authenticated employee access

    The tool sits behind an auth layer so only the client staff can generate documents.

  • Async download flow

    Generation and document export are decoupled (separate summaryId-based steps), so the operator gets fast feedback that generation succeeded before the heavier document assembly runs.

The Outcome

What previously took the client team hours of manual writing per course now happens in minutes. The product has been in active production use through three rounds of customer feedback, with the team continuing to refine prompts and output formatting based on real student-facing requirements. The success of this product directly led to the client commissioning a second tool from us — see the Practice Exam Generator case study.

Tech Stack

Next.jsNode.jsOpenAI APIcustom Word document generation pipelineLaTeX rendering layer

Book a Discovery Call

Tell us what you are building. We will recommend an engagement model and honest next steps.

Book a Discovery Call