Agentic AI Developer - Orange Tree Systems

We're hiring an [b]Agentic AI Developer[/b] to build production-grade [b]AI calling agents[/b] — voice-first autonomous agents that make and receive phone calls, hold natural conversations, and complete real tasks end-to-end. This is a hands-on role for someone who has already shipped calling agents to real users and knows the difference between a demo and a system that handles thousands of calls without breaking. If you've built voice agents on [b]Vapi, Retell AI, Bland AI, Synthflow, or LiveKit[/b], or scripted custom flows using [b]Twilio + Deepgram + ElevenLabs + an LLM[/b], we want to talk. [h3]What You'll Do[/h3] [ml][ul][li indent=0 align=left]Build AI calling agents (inbound and outbound) for real use cases — booking, lead qualification, support, follow-ups[/li][li indent=0 align=left]Design conversation flows, reasoning loops, tool use, and human handoff logic[/li][li indent=0 align=left]Integrate agents with CRMs, calendars, databases, and third-party APIs[/li][li indent=0 align=left]Optimize for the three things that matter in voice: [b]latency, accuracy, and natural conversation[/b][/li][li indent=0 align=left]Handle real-world challenges — interruptions, background noise, accents, silence, call transfers[/li][li indent=0 align=left]Monitor, debug, and continuously improve agents in production[/li][/ul][/ml][h3]What You Must Have[/h3] [ml][ul][li indent=0 align=left][b]Proven experience building calling agents[/b] — via scripting (Python/Node.js) on telephony APIs, or using platforms like Vapi, Retell, Bland, or Synthflow[/li][li indent=0 align=left]Hands-on work with [b]LLMs[/b] (Claude, GPT, Gemini) — prompt engineering, tool calling, context management[/li][li indent=0 align=left]Strong grasp of the [b]voice AI stack[/b]: STT (Deepgram, AssemblyAI), TTS (ElevenLabs, Cartesia), telephony (Twilio, Plivo)[/li][li indent=0 align=left]Solid [b]Python or Node.js[/b] skills[/li][li indent=0 align=left]Understanding of [b]latency optimization[/b] — you know why 1.5s response time kills a call, and how to get it under 800ms[/li][li indent=0 align=left]Ability to design flows that handle real human messiness: interruptions, off-topic chatter, unclear requests[/li][/ul][/ml][h3]Nice to Have[/h3] [ml][ul][li indent=0 align=left]Experience with [b]LangGraph, CrewAI, or AutoGen[/b][/li][li indent=0 align=left]Familiarity with [b]MCP[/b] for tool integration[/li][li indent=0 align=left]Exposure to [b]RAG[/b] and vector databases[/li][li indent=0 align=left]Compliance awareness — [b]TCPA, HIPAA[/b], or similar[/li][li indent=0 align=left]Domain experience in healthcare, fintech, or real estate[/li][/ul][/ml]