"OncoAgent: A Dual-Tier Multi-Agent Framework for Privacy-Preserving Oncology Clinical Decision Support"

OncoAgent combines a dual-tier fine-tuned LLM architecture with a multi-agent LangGraph topology. The system routes clinical queries through an additive complexity scorer to either a 9B parameter speed-optimized model (Tier 1) or a 27B deep-reasoning model (Tier 2). Both models were fine-tuned via QLoRA on a corpus of 266,854 real and synthetically generated oncological cases using the Unsloth framework on AMD Instinct MI300X hardware (192 GB HBM3). Sequence packing on MI300X enabled full-dataset fine-tuning in approximately 50 minutes — a 56× throughput acceleration over API-based generation. The system also implements a four-stage Corrective RAG pipeline over 70+ physician-grade NCCN and ESMO guidelines.

Safety is enforced via a three-layer reflexion safety validator that maintains a strict Zero-PHI policy—no protected health information ever leaves the hospital network. Post-fix, CRAG document grading achieved a 100% success rate with a mean RAG confidence score of 2.3+. The complete system is 100% open source and deployable on-premises, eliminating proprietary cloud API dependency and preserving patient data sovereignty. By decomposing clinical reasoning across eight specialized LangGraph nodes, OncoAgent ensures auditable, hallucination-grounded recommendations suitable for high-stakes oncology triage and treatment pathway decisions.