All roles

Software Engineer - AI Agents

Engineering Remote (US) Full-time

Build the agent infrastructure — LLM orchestration, real-time voice/video, and the tools our AI employees use to get real work done.

We're hiring a Software Engineer - AI Agents to help build the core systems that power Vigy's AI employees: LLM-driven agents that take real actions across email, calendars, CRMs, and dozens of other tools. You'll work across our Rails backend, real-time voice/video pipeline (LiveKit + OpenAI Realtime), and the integrations layer that lets our agents act on a customer's behalf. This role is for someone who wants to ship production AI systems, not just prototype them.

What you'll do

  • – Design and build agentic workflows on top of OpenAI's API (Realtime, function/tool calling, structured outputs)
  • – Develop and maintain the Rails backend that powers agent conversations, job execution, and integrations
  • – Improve the real-time voice/video pipeline (LiveKit, HeyGen LiveAvatar) for latency and reliability
  • – Build and maintain tool integrations (via Composio and direct APIs) that let agents take real actions
  • – Write background jobs (SolidQueue) that handle long-running or asynchronous agent work
  • – Debug production issues across the LLM, backend, and real-time layers, distinguishing transient failures from real bugs
  • – Collaborate closely with founders on product direction for what AI employees should be able to do next

What we're looking for

  • – 3+ years of professional software engineering experience, ideally with some in Ruby on Rails
  • – Hands-on experience building with LLM APIs (OpenAI, Anthropic, or similar) — prompting, tool/function calling, structured outputs
  • – Comfortable working across the stack: backend services, background jobs, and third-party API integrations
  • – Experience with or strong interest in real-time systems (WebSockets, streaming audio/video) is a plus
  • – Product sense — you can reason about what makes an AI employee actually useful, not just technically correct
  • – Pragmatic engineering judgment: you ship, measure, and iterate rather than over-engineering up front
Apply for this role