Back to projects

Smart Digitech

Pipecat Voice AI Agent

Realtime voice AI agent using Pipecat with Plivo transport, WebRTC, and flexible model backends including LLM APIs, Gemini Live, OpenAI Realtime, and vLLM.

PipecatPlivoWebRTCOpenAI RealtimeGemini LivevLLMHugging FaceLLM APIs

Responsibilities

  • Built a voice AI agent flow around Pipecat with Plivo transport and WebRTC realtime communication.
  • Designed the agent to support multiple model modes, including standard LLM usage, Gemini Live, OpenAI Realtime, and local/self-hosted vLLM backends.
  • Set up vLLM infrastructure and evaluated Hugging Face models for voice-agent use cases and backend compatibility.

Outcomes

  • Created a flexible voice AI foundation that can switch between hosted realtime APIs and self-hosted model serving.
  • Improved experimentation speed for comparing model quality, latency, and deployment trade-offs.

Need similar work?

I am open for freelance projects involving product builds, integrations, automation, and platform work.

Start a Project