An Automated Retrieval-Augmented Generation LLaMA-4 109B-based System for Evaluating Radiotherapy Treatment Plans

This project occupies a highly specialized intersection of medical physics and NLP. The use of a 100B+ parameter model (cited as LLaMA-4, likely LLaMA-3 or a future-dated placeholder) for radiotherapy plan evaluation is a novel application of RAG. The defensibility is currently a 4 because while the domain expertise required to curate the 614-plan dataset and normalize dose metrics is significant, the underlying RAG architecture is standard. The lack of stars (0) indicates no developer traction, though the 18 forks suggest internal research interest or academic replication. The real moat in this space isn't the code—it's the regulatory clearance (FDA/CE) and the integration with treatment planning systems (TPS) like Varian's Eclipse or RaySearch's RayStation. Frontier labs like OpenAI are unlikely to enter this niche due to high liability and low market volume. The primary threat comes from established radiotherapy software vendors who will likely integrate similar LLM features into their existing, proprietary clinical suites within the next 24 months.