Wood AI

My startup cofounder, Jivin, and I saw this project from idea to minimum viable product over the course of 2 months. An in-depth explanation of the product and its use cases is on the project website, woodai.us.

My work focused on the backend, built using Python Google Cloud Functions and Firestore. Most of the technical challenge was to dynamically collect context for each user prompt and return it quickly. One particular challenge was to resolve long user prompts without exceeding Google Cloud's one minute timeout for Cloud Run Functions. I used a job orchestration system to repeatedly call the same wrapper function, storing collected data along the way, until the final response was ready to be returned.