UnimatchLab Neybo 3 дня назад
Senior LLM Engineer (Voice AI)
- Удалённо
- Опыт от 4 лет
- Полная занятость
PythonAIDockerLLMMLremoteRAGWebsocketHiringVoice AITTSWebRTCSTTAsyncPipecatAIEngineer
Описание
Responsibilities
- Manage a cloud voice pipeline: end-of-phrase detection, STT, LLM, TTS with an end-to-end latency of 1-2 seconds (p50 and p90).
- Manage the streaming pipeline: barge-in, streaming tokens to TTS, timeouts, fallback between providers.
- Adjust end-of-phrase detection for long pauses from children (silence threshold 1.5-3 s).
- Build an evaluation stand on children's speech with automatic runs.
- Build a security cascade with rule updates without firmware releases.
- Build responses on verified content (RAG, templates, tool calling).
- Configure tracing, session cost, and zero retention with providers.
Requirements
- 4+ years in ML/LLM, including 1.5+ years with production voice agents.
- Streaming cascades with barge-in and fallback; Pipecat, LiveKit Agents or similar.
- Latency optimization by percentiles; WER, TTFT; unit economics of STT/TTS.
- Guardrails and security classification for sensitive audiences.
- Python, async, WebSocket/WebRTC, Docker, cloud; English B2+.
Nice to have
- Children's and edtech products; COPPA, PRIVO, kidSAFE.
- Children's or accented speech, Arabic in STT/TTS.
- AEC and audio streaming from Linux devices; speech-to-speech.
We offer
- Remote work, full-time employment.
- Location: Europe, MENA, LATAM.