01
Design an AI agent within FlutterFlow's Agent Builder capable of understanding complex medical data — vitals, symptoms, and patient history — and delivering clinically reliable interpretations, where generic AI responses were not acceptable for emergency
Created a custom AI agent in FlutterFlow with carefully engineered system prompts, medical terminology rules, and validated example conversations, enabling accurate and consistent analysis of SAMU patient records directly inside the app.
02
Force the AI agent to always respond in a strict, structured format — free-form AI text could not be used for clinical handoffs, where a missing field could compromise patient safety.
Configured the FlutterFlow AI agent with JSON response mode and a locked SBAR schema (Situation, Background, Assessment, Recommendation), ensuring every generated report is complete, consistent, and ready for emergency department handoff.
03
Handle AI agent calls, vital sign monitoring, and live data updates simultaneously without freezing the interface — paramedics cannot wait on a loading screen during an emergency.
Implemented asynchronous AI agent actions in FlutterFlow with loading states, response caching, and optimized action flows, keeping the app fully responsive while AI analysis runs in the background.
04
Pass dynamic context — current vitals, mission details, and patient history stored in the backend — into the AI agent's prompts at runtime so every analysis is personalized to the active case.
Built dynamic prompt injection in FlutterFlow that merges real-time patient data and mission logs into the agent's context on every call, producing case-specific analysis and alerts instead of generic responses.
05
Transform the AI agent's structured responses, mission logs, and vitals data into professionally formatted PDF reports that meet medical record-keeping requirements.
Developed a PDF export pipeline that parses the agent's JSON output and compiles it with mission data into standardized, shareable reports — completing the workflow from AI analysis to official documentation.
06
Prevent hallucinated or clinically irrelevant AI suggestions by constraining the agent to real-world SAMU emergency protocols and field terminology.
Refined the FlutterFlow AI agent's system prompts and example conversations in collaboration with an experienced SAMU paramedic, aligning every output with real emergency response workflows and the high-pressure environment of field medicine.
