SylabiTechnology
A real intelligence stack, built for the classroom.
Sylabi combines document AI, a curriculum knowledge graph, and grounded language models — engineered for African university networks, not Silicon Valley demo days.
The pipeline
From raw PDF to a study plan you can trust.
1. Ingest
Students and faculties upload PDFs, DOCX outlines or paste raw text. We also crawl publicly-listed course pages with permission.
2. Parse
A layout-aware parser splits every syllabus into topics, sub-topics, learning outcomes, assessment weights and reading lists.
3. Normalise
Topics are mapped to a shared taxonomy so 'Ohm's Law' at UNILAG equals 'Ohm's Law' at KNUST — even when the wording differs.
4. Diff
A semantic diff engine finds overlaps, gaps and weight shifts across any two syllabi in under a second.
5. Link
Each topic is joined with past-question banks, textbook chapters and vetted online lectures via a curriculum knowledge graph.
6. Explain
An LLM layer summarises differences and suggests a study plan — always grounded in cited source syllabi.
The knowledge graph
Every topic, course and resource — one connected graph.
Topics, lecturers, past questions and reading lists are modelled as nodes in a queryable graph, so a diff isn't just text matching — it's a walk across shared curriculum structure.
The stack
Built on tools we can operate — not ones we can only demo.
Parsing
Layout-aware PDF/DOCX pipeline with OCR fallback for scanned handouts.
Knowledge graph
Topics, courses, lecturers and resources modelled as a queryable graph.
Vector search
Embeddings power semantic matching across differently-worded outlines.
GPU inference
NVIDIA GPUs run our embedding and LLM inference, keeping diffs fast even at scale.
Cloud infrastructure
Built on AWS — edge functions and managed compute keep the app fast on 3G and cheap to scale per student.
Postgres + storage
Every syllabus versioned, every diff reproducible, every source cited.
Safety & privacy
Uploads are scoped to your school. Personal data never leaves the region.
Want a technical deep-dive?
We're happy to walk research teams and university IT through the architecture.