Real-Time Voice AI Engineer

Wati; Hong Kong; On-site; HKD 700,000 - 1,200,000; Full time

Job summary

Wati is seeking an Agentic Engineer - Voice AI to build and scale real-time voice AI capabilities on WhatsApp. You will design systems that listen, think, and speak in real time, integrating ASR, TTS, and LLMs to deliver fluent conversations.

You'll work on latency-sensitive infrastructure, from audio transport to model inference, and help orchestrate tool calls and multi-turn dialogue across its AI agent stack for 190+ countries.

Qualifications

• 3+ years of software engineering with backend focus (Go or Python).
• Experience with real-time communication: WebRTC, RTP/RTCP, audio codecs.
• Familiarity with AI/LLM integration: APIs, tool calling, orchestration.
• Interest in speech tech: ASR, TTS, audio processing pipelines.
• Understanding distributed systems and cloud-native architectures (GCP).

Responsibilities

• Design and optimize real-time voice AI pipelines from transport to inference and synthesis.
• Integrate frontier AI models and cascade architectures (ASR → LLM → TTS).
• Maintain media infrastructure: LiveKit routing, Opus codec handling, RTP/RTCP, and voice activity detection.
• Develop voice interaction capabilities: tool calls, context engineering, multi-turn management.
• Reduce end-to-end latency from audio capture to AI response playback.
• Collaborate with product and platform teams to deliver production-grade voice experiences on WhatsApp.
• Ensure reliability, performance, and scalability for customers across 190+ countries.

Skills

Go

Python

WebRTC

AI/LLM integration

Speech technologies

Distributed systems

PostgreSQL

Redis

Pub/Sub

Debugging latency

Tools

LiveKit

OpenAI Realtime API

Google Gemini Live

LangChain

Dify

CrewAI

GCP

Job description

Wati is seeking an Agentic Engineer - Voice AI to build and scale real-time voice AI capabilities on WhatsApp. You will design systems that listen, think, and speak in real time, integrating ASR, TTS, and LLMs to deliver fluent conversations.

You'll work on latency-sensitive infrastructure, from audio transport to model inference, and help orchestrate tool calls and multi-turn dialogue across its AI agent stack for 190+ countries.