About the Role
We’re looking for a hands-on engineer to build real-time, voice-enabled autonomous AI agents for next-generation Sales CRM workflows.
You’ll combine low-latency voice streaming, stateful multi-agent orchestration, and production-grade backend engineering to ship agents that can hold natural conversations and act inside a CRM.
What You’ll Do
- Build conversational voice agents using LiveKit Agents, Pipecat, or similar real-time frameworks
- Design and tune STT and TTS pipelines for low latency and natural-sounding conversation
- Engineer real-time audio streaming over WebSockets/WebRTC, including VAD, dynamic interruption handling, and turn-taking logic
- Build stateful, multi-agent workflows with tool calling using LangGraph and LangChain
- Integrate multiple LLM providers (OpenAI, Anthropic Claude, Google Gemini)
- Use graph-based architectures, graph databases, and GraphRAG for knowledge retrieval and contextual grounding
- Build and integrate REST APIs, and ship containerized services with tracing and evaluation in place
Required Skills
- Strong Python with robust asynchronous programming (asyncio) and scalable backend design
- Hands-on experience with LiveKit Agents, Pipecat, or equivalent voice agent frameworks
- STT experience: Deepgram, Whisper, AssemblyAI, or equivalent
- TTS experience: ElevenLabs, OpenAI TTS, Azure Speech, or equivalent
- Real-time audio streaming (WebSockets/WebRTC), low-latency pipelines, VAD, interruption handling
- LangGraph and LangChain for stateful orchestration, multi-agent collaboration, and tool calling
- Multi-provider LLM integration (OpenAI, Claude, Gemini)
- Graph databases and GraphRAG
- Observability and tracing (OpenTelemetry, LangSmith), LLM evaluation frameworks, Docker/containerized deployment
Rate:
Not specified
Location:
United States
IR35 Status:
Outside
Remote Status:
Remote
Industry:
AI & Machine Learning
Seniority Level:
Not Specified