MedNav Agent
An empathetic, plain-language patient navigator that translates complex medical terminology and clinical trial records into friendly, clear, and easy-to-understand explanations.
Project Overview
A secure, multi-agent AI assistant built using the Google ADK (Agent Development Kit) 2.0. It orchestrates specialized sub-agents (MedJargonAgent and ClinicalTrialAgent) via local Model Context Protocol (MCP) server tools to safely explain medical terms and query clinical trials, guarded by pre-processing security checkpoints and human-in-the-loop validation.
The Problem
Patients often struggle to navigate the healthcare system due to complex medical jargon and dense clinical trial documentations. Additionally, deploying navigational medical AI requires strict safety safeguards to prevent data leaks (PII) and unsafe diagnostic advice.
The Solution
Implemented an ADK 2.0 workflow graph with a Security Checkpoint (performing PII scrubbing, prompt injection defense, and medical safety checks) routing to a MedNav Orchestrator Agent. The orchestrator delegates tasks to specialized sub-agents running FastMCP tools, and terminates with a human review node for HITL approval.
The Impact
Bridges the communication gap between healthcare providers, clinical researchers, and patients. By transforming intimidating clinical trial pages and medical terminology into simple, reassuring text, it empowers patients to make informed decisions about their care, increases enrollment in clinical trials, and reduces patient anxiety.
Date / Year
2026