Decurb – AI Token Forecasting & Optimization Platform

Project Details

Decurb is an AI platform I’m currently leading as part of a collaborative team project, aimed at helping users better understand and optimize how much “token space” their prompts and completions consume when interacting with large language models like GPT. The platform analyzes token usage in real-time, forecasts when prompts are likely to exceed model limits or produce incomplete responses, and proactively manages those cases by breaking complex prompts into smaller subproblems. It also caches reusable outputs and recombines them intelligently to reduce cost while preserving accuracy and performance. Decurb was born out of both technical curiosity and a practical need for scalable, efficient AI workflows—especially in applications that involve long or dynamic user interactions. Stay Tuned!