Buildspace AI is a project aimed at fostering connections and collaborations among Buildspace members and other builders through AI-powered voice conversations. The goal was to create a seamless platform that allows users to find each other and exchange knowledge.

I implemented a comprehensive data pipeline that involved web scraping participant information and video submissions from Buildspace Season 3 demo day submissions, transcribing over 300 videos using the Deepgram transcription model, and embedding the transcriptions using OpenAI's ADA model.

The embedded data was stored in a PostgreSQL database hosted on Supabase, enabling efficient retrieval and analysis.

I leveraged the OpenAI's GPT-4 model to generate contextually relevant responses based on user input during live voice conversations. To enhance the user experience, I integrated Elevenlabs' text-to-speech functionality, allowing the AI to communicate through natural-sounding voice responses.

you can checkout the live demo here :

https://buildspace.vic.so