PODSAGE

PODSAGE is an AI-driven educational storytelling platform that transforms complex concepts into engaging, interactive narratives. By combining fine-tuned large language models with multimodal delivery, PODSAGE enhances comprehension and long-term retention, particularly for neurodivergent and diverse learners (Ben seddik, 2025).

The platform generates adaptive, citation-aware stories that integrate text, audio, and visuals, grounded in cognitive science and educational psychology.

Technologies: Ollama (LLaMA 3.3), PyTorch, RAG, ElevenLabs, DeepFloyd IF, Flutter, FastAPI, Django, Spring Boot, Node.js, PostgreSQL, MongoDB, Redis, Google Cloud Platform

Main Contributions

• A scalable, user-centric framework for generating emotionally engaging and pedagogically effective stories from complex concepts using LLMs.
• A demonstration that fine-tuning a lightweight LLM on a curated dataset distilled from a more powerful model (GPT-4.1) enables high-quality, efficient story generation.
• A platform supporting both human and LLM-based evaluation, enabling continuous improvement and trust in educational storytelling systems.

Research Directions Enabled by PODSAGE

• Psychographic, Behavioral, and Cognitive Data Collection
  Longitudinal analysis of user interaction data (choices, response times, quiz performance) to study how narrative structure and personalization impact learning and memory.

• Human Subject Studies
  Controlled experiments comparing narrative techniques, formats, and feedback mechanisms to rigorously evaluate retention and comprehension.

• LLMs as Educational Evaluators
  Use of LLMs as scalable, consistent evaluators of narrative quality, educational value, and learning objective alignment.

• Narratives for LLM Reasoning
  Investigation of whether story-based prompting improves LLM reasoning, problem-solving, and knowledge transfer in complex domains.