We aim to remove access to data as the principal roadblock when deploying deep learning in pharmaceutical R&D.

Why does this matter?

By minimizing the constraints of data collection, we aim to vastly expand the applicability of deep learning in drug discovery.

Imagine the capability to model lower-throughput biological assays that more faithfully recapitulate the complexities of human disease.

Technology Overview

Novel deep learning approaches for de novo drug design in low data environments

Technologies in Development

Clinical development

Low data structure-activity prediction

Few-shot learning algorithms capable of learning accurate structure-activity relationships in extreme low-data settings.

Preclinical toxicology

De novo drug design

Generative algorithms for molecular design in a fully automated fashion without the need for human intervention.

Drug-Drug interactions

Synergistic drug-drug interactions

Deep learning algorithms to enable efficient identification of synergistic compound pairs.


Retrosynthesis planning

Deep learning algorithms for accurate computer-aided retrosynthesis.

Explainer Functionality

Active learning algorithms for in-silico-in-vitro feedback loops to enable improved predictive models and more efficacious compounds.

Enhancer Functionality

Reinforcement learning algorithms for the parallel optimization of multiple physicochemical properties.

E 2

Augment medicinal chemistry by allowing for rapid converge towards lead-like compounds with minimal human intervention.