AIML-powered Low-cost, High-octane Eco-fuel Mixture Identification Strategy & Toolkit
As part of the SHELL.ai hackathon of 2025, we developed the ALCHEMIST framework to predict chemical properties of synthetically blended fuels. The entire codebase is available on GitHub. The associated documentation can be found here. The trained model weights can be downloaded from Hugging Face. The original ALCHEMIST framework is comprehensive and makes use of several state-of-the-art AIML models: AutoGluon, RealMLP, and TabPFN.
Here, we present ALCHEMIST Jr., a lightweight version of the original framework designed for quick experimentation. This interactive implementation enables users to make fuel blend property predictions in real-time without the computational complexity of the full ALCHEMIST framework.
