Voice AI Engineer

NAIRON GmbH; United States; Remote; USD 140,000 - 210,000; Full time

Benefits offered by this job

Remote work

Laptop provided

Phone provided

Job summary

NAIRON GmbH is seeking a Senior/strong mid-level engineer to optimize real-time voice conversations across streaming ASR, LLM processing, and TTS. You will work on production audio systems, latency reduction, and model serving on GPUs, with ownership from diagnosis to production fixes.

You will collaborate across a compact team, using Python with asyncio, and apply measurement-driven improvements. Remote role with US-wide flexibility.

Qualifications

• Senior or strong mid-level experience in production real-time audio, conversational AI or machine-learning model serving.
• Strong Python skills, including asyncio and streaming pipelines.
• Hands-on experience with speech technologies such as STT, TTS or VAD, and LLM APIs in latency-sensitive systems.
• Experience serving models on GPUs, with an understanding of batching and memory constraints.
• English proficiency for written and spoken work.
• A measurement-led approach and ownership from diagnosing a poor call to delivering a production fix.

Responsibilities

• Improve conversation turn-taking, including voice activity detection, end-of-turn detection and handling interruptions.
• Reduce end-to-end latency across speech recognition, language-model processing and speech synthesis, including smooth provider failover.
• Improve conversational LLM use through prompting, tool calls, streaming and parallel or speculative requests.
• Serve speech models on GPUs and measure throughput and time to first audio.
• Fine-tune speech or language models for customer languages and voices.
• Build test calls, benchmarks and quality metrics to evaluate changes in real conditions.

Skills

Real-time audio

Python (asyncio)

STT/TTS/VAD

LLM APIs latency

GPU model serving

English proficiency

Job description

Role overview

Improve the speed, naturalness and reliability of voice conversations in live calls. This role works across streaming speech recognition, language models and speech synthesis, combining production engineering with measurement and model-serving work.

Responsibilities

• Improve conversation turn-taking, including voice activity detection, end-of-turn detection and handling interruptions.
• Reduce end-to-end latency across speech recognition, language-model processing and speech synthesis, including smooth provider failover.
• Improve conversational LLM use through prompting, tool calls, streaming and parallel or speculative requests.
• Serve speech models on GPUs and measure throughput and time to first audio.
• Fine-tune speech or language models for customer languages and voices.
• Build test calls, benchmarks and quality metrics to evaluate changes in real conditions.

Requirements

• Senior or strong mid-level experience in production real-time audio, conversational AI or machine-learning model serving.
• Strong Python skills, including asyncio and streaming pipelines.
• Hands-on experience with speech technologies such as STT, TTS or VAD, and LLM APIs in latency-sensitive systems.
• Experience serving models on GPUs, with an understanding of batching and memory constraints.
• English proficiency for written and spoken work.
• A measurement-led approach and ownership from diagnosing a poor call to delivering a production fix.

Nice to have

• LiveKit Agents, Pipecat or WebRTC.
• Speech or language model fine-tuning, including LoRA, and audio processing such as echo cancellation or noise suppression.
• German language skills or experience with German speech data.

Benefits and work setup

• Work remotely or from a co-working space, with a laptop and phone provided.
• Join a small, agile team and work across the voice stack, from model serving to the caller’s experience.