voice agents · LLMs · evaluation · teamwork
KermitPanic · Voice Agents
A voice agent to manage medical appointments, notify the corresponding doctor and keep track of everything that happens. Project developed for HackSpain.
- Role
- Evaluation harness and integration
- Period
- 18-20 September 2026
- Status
- shipped
- Visibility
- public repository
The challenge
At HackSpain 2026, my team built a voice agent to handle calls and manage appointments in a fictional clinic. The agent had to converse with the patient and execute the operation against the Prosper AI challenge platform.
My contribution
I built the evaluation environment, a scenario catalogue and a comparator that verifies each field of the final booking. This let us test changes without consuming the official platform’s limited minutes and evaluate our agent’s performance.
Screenshots
