AI Voice Engineering & Product Intern

Overview

At Skyline, a Berkeley SkyDeck-backed startup in commercial real estate, I built a real-time AI voice agent designed to automate introductory client qualification calls for CRE brokers. I owned both the engineering and product sides: defining the STT/LLM/TTS pipeline architecture, evaluating competing voice AI models, and building the React frontend. The system targets 7–10 minute qualification calls and client profile creation, saving brokers up to 1.5 hours of manual call time per day.

Responsibilities

• Built a real-time AI voice agent pipeline for automated CRE broker qualification calls.
• Defined the Speech-to-Text/LLM/TTS architecture and evaluated ElevenLabs, Deepgram, Whisper, and Gemini to inform stack selection.
• Developed the React UI frontend with Tailwind CSS and shadcn/ui, implementing Listening, Thinking, and Speaking states.
• Owned product requirements and system design from initial scoping through implementation.

Achievements

• Designed a voice agent system targeting up to 1.5 hours of saved call time per broker per day.
• Evaluated four competing voice AI providers to make a data-informed stack selection.
• Delivered a full-stack working prototype spanning pipeline architecture and a polished React frontend.

Skills