Phlow, Enveda apply AI-driven chemistry to accelerate API development

Phlow, a U.S.-based contract development and manufacturing organization (CDMO) headquartered in Richmond, Virginia, and Enveda, a biotechnology company focused on AI-driven drug discovery, have demonstrated a new approach to active pharmaceutical ingredient (API) process development that uses internally generated reaction data to improve AI-based predictions of reaction performance.

The companies said a joint pilot program launched last year generated nearly 20,000 unique chemical reactions in three months, creating a large, uniform dataset used to train AI models for predicting reaction kinetics, yields, and purity profiles. According to the companies, the approach significantly improved prediction accuracy compared with models trained on publicly available datasets alone, reducing development timelines from years to months.

Phlow and Enveda said the work compared traditional machine learning methods with deep learning models, including graph neural networks that represent molecules as connected systems of atoms and bonds. The companies contend this modeling approach more closely reflects real-world chemistry and delivers stronger predictive performance for reaction optimization and future retrosynthetic planning.

According to the companies, the ability to rapidly generate and integrate high-quality experimental data into AI models reduces reliance on trial-and-error experimentation, supports faster route selection, and improves scalability for drug substance manufacturing. The collaboration is intended to support more efficient development of essential medicines while strengthening domestic manufacturing capabilities.

Phlow said it plans to expand the framework to additional reaction classes and apply deep learning to a broader retrosynthetic analysis platform, enabling faster and more sustainable chemical synthesis across a wider range of applications.

The AI-driven chemistry effort builds on Phlow’s broader investments in U.S.-based pharmaceutical manufacturing. In 2025, the company raised $37 million to expand its cGMP manufacturing footprint and scale AI-powered development systems, according to the company. 

Phlow said it is operating two U.S. facilities capable of producing APIs at kilogram and metric-ton scale using batch and continuous processes and has expanded research labs to accelerate small molecule development and scale-up.