EXPLORE

Mnemosyne: academic synthesis & exam intelligence.

An autonomous academic synthesis engine analyzing historical university question repositories to predict question frequencies, isolate syllabus dependencies, and synthesize high-yield revision dossiers.

View GitHub Repository ↗
  • LLM Synthesis
  • Vector Embeddings
  • PDF Document Pipeline
  • Deterministic RAG

The Engine.

University students and researchers face massive repositories of unstructured exam papers, lecture notes, and textbook PDFs. Mnemosyne parses dense technical literature through an automated multi-stage ingestion pipeline.

By extracting semantic embeddings from curriculum syllabi and pairing them with frequency matrices of past examination prompts, Mnemosyne builds an evolving knowledge graph that flags high-probability concepts, identifies prerequisite gaps, and renders structured Markdown revision manuals.

Mnemosyne academic synthesis engine user interface and knowledge cluster

The Bottleneck, The Solution.

Bottleneck

Hallucination in Generic Language Models

Off-the-shelf generative models frequently hallucinate syllabus requirements, invent non-existent course topics, and fail to calculate historical topic distribution weights.

Breakthrough

Grounded Vector Retrieval & Frequency Scoring

Every synthesized output is strictly bounded by deterministic RAG pipelines, linking every summarized concept back to exact textbook page coordinates and past exam session IDs.

Under The Hood.

Python 3.11 FastAPI ChromaDB Sentence Transformers PyMuPDF / pdfplumber React & Vite UI
0
Topic Accuracy (%)
0
Pages Processed/Min
0
Hallucination Rate (%)
0
Lighthouse Performance
Next project: KAGE Agent
Next Research Project ↗

KAGE Agent: proactive vulnerability
triage & threat attribution.