InternshipInternshipsApr 2026 – Jun 2026·Philadelphia, PA · Penn Innovation Award Winner
Stealth AI Startup
AI Engineering Intern
Instrumented and sped up a latency-sensitive conversational AI pipeline spanning Next.js, Pipecat Cloud, Daily WebRTC, Tavus, and LLM/TTS services.
Executive summary
AI Engineering Intern at a stealth AI startup in Philadelphia (Penn Innovation Award Winner). The work focused on a live conversational avatar stack where startup latency and session reliability mattered in production. I profiled the critical path across Next.js, Pipecat Cloud, Daily WebRTC, Tavus, and LLM/TTS services, then redesigned session init so the avatar and first speech came up faster and more reliably.
- Instrumented and profiled a latency-sensitive conversational AI pipeline spanning Next.js, Pipecat Cloud, Daily WebRTC, Tavus, and LLM/TTS services to isolate critical-path startup bottlenecks.
- Redesigned session initialization through asynchronous preconnection, connection-state synchronization, and WebRTC sequencing.
- Cut post-countdown avatar load to ~5s and first speech to ~7s, and improved production reliability.
Next.jsPipecat CloudDaily WebRTCTavusLLM/TTSLatency
Proof
Pictures and demos from the role
Visual evidence first; deeper engineering write-ups live on the related project pages.