Better predictions from limited scientific data -- coming soon.
A new neural-network architecture for accurate and reliable predictions when data is scarce.
LeafSplit is developing a new neural network architecture designed for data constrained scientific problems. Our technology aims to help AI models extract more useful signal from limited, expensive training data where conventional neural networks can be inaccurate, unstable, or prone to overfitting.
We are initially validating the architecture for molecular prediction and drug discovery, with potential applications in medical research, materials science, aerospace, and space fields where experiments are costly and high-quality data is scarce.
Make AI models that extract more useful signal from small, expensive scientific datasets.
Enable earlier, more reliable predictions for drug discovery, medicine, materials science, aerospace, and space.
Develop and validate an architecture that improves accuracy, stability, and generalization under real world data constraints.
Join LeafSplit’s early community of researchers, AI teams, industry partners, and supporters working to unlock discovery from limited scientific data.
We’d love to hear from you. Reach out anytime or follow us for updates.