1. AI providers used
- Google Gemini Live API - Powers the real-time voice agent. Handles voice activity detection (VAD), speech-to-text (STT), language model reasoning (LLM), and text-to-speech (TTS) natively within each practice session. Session audio is transmitted in real-time and is not persistently stored by Visaprep Agent.
- Google Gemini API - Used server-side for post-session coaching analysis, document content extraction, and visa intelligence data refresh. Text inputs (transcripts, documents) are processed but not used to train Google's models under standard API terms.
- LiveKit Cloud - Provides the WebRTC voice transport and real-time agent orchestration infrastructure. Routes audio between your device and the AI voice agent. LiveKit does not perform AI inference; it provides the communication layer.
2. Current principles
- AI is used for interview practice, coaching responses, summarization, preparation feedback, and related product workflows.
- AI output is not legal advice, immigration advice, eligibility analysis, or an outcome prediction.
- User interview transcripts, session audio, uploaded documents, and prompts are not used to train AI models by Visaprep Agent.
3. Human review
Users must review AI output before relying on it. Official government instructions and qualified legal advice control over AI-generated suggestions.
Complex legal questions should be taken to a licensed immigration attorney or accredited representative.