Shift Hackathon - Gen AI

During the Shift GenAI Hackathon, I pitched Pimpon, a smart voice assistant for emergency call handling. Out of over 40 ideas, mine was selected in the top 10. I led a team of 6 people to build the project.

Pimpon handles calls before medical dispatchers (ARM) step in. It gathers key information using LLMs and function calling to reduce call handling time.

We used Apple Shortcut for voice simulation and FastAPI + WebSocket for real-time backend. The assistant interacts with GPT‑4o, enhanced with RAG and custom functions.

The project won 3rd place, praised for its social impact, execution, and deployment potential.

• Pitched in front of 70+ people, selected among top 10 ideas.
• Formed and led a 6-person team throughout the weekend.
• Developed Pimpon, an AI medical voice assistant using GPT‑4o, RAG, and function calling.
• Used Apple Shortcut and TTS to simulate user calls.
• Built a real-time app using WebSocket and FastAPI backend.
• Awarded 3rd place in Shift GenAI Hackathon.

Simulated Call Demonstration

We scripted a call to the Pimpon assistant to showcase how it works. The caller interacts vocally with the assistant, which asks key questions to assess the medical situation before handing over to a medical dispatcher (ARM). This highlights the assistant’s value in offloading emergency lines.

Advanced Demonstration via UI

This demo shows a complete interaction with the Samuel assistant through a graphical interface. I simulate a caller interacting with the assistant. When Samuel detects an emergency, it automatically escalates the call to an ARM. Afterward, the ARM can view a detailed summary of the exchange to support effective care.