Capstone — VivaAgent#

Objective#

Create a real-time, voice-based AI agent capable of conducting an oral examination (viva) for students.

Tech Stack#

  • FastAPI + WebRTC: For real-time audio streaming.
  • Claude 3.5 Sonnet: For conversational logic and grading.
  • Whisper & ElevenLabs: For STT and TTS.
  • LangGraph: To manage the state of the exam (Introduction -> Questioning -> Grading).

Requirements#

  1. The agent must ask 3 technical questions based on a provided syllabus.
  2. It must listen to the student’s answer, ask follow-up questions if the answer is vague, and grade the response.
  3. Latency must be kept under 1.5 seconds.

Deliverables#

  • A working web interface.
  • Grading transcripts showing the AI’s reasoning.