ChatToStory
ChatToStory is an end-to-end pipeline for LLM-based story generation. The system converts conversational input into coherent narratives by combining large language models with a lightweight, fine-tuned model for task-specialized inference.
I led the team and worked across the full pipeline, which combines:
- Large language models including NVIDIA Nemotron, GPT, Gemma, Llama, and Mistral for narrative generation.
- Fine-tuned T5 using Hugging Face Transformers as a lightweight, task-specialized inference solution.
- Automated parallel GPU workloads on a Linux HPC cluster using SLURM and Bash for data partitioning, resource allocation, and batch processing.
- End-to-end data processing from raw conversational input to generated narrative output.
Project link: https://github.com/RaedFidawi/ChatToStory