Draft:Evan Feinberg

Summary

Evan Feinberg, Ph.D., is an American ML researcher, biophysicist, entrepreneur, and the founder and CEO of Genesis Molecular AI (formerly Genesis Therapeutics), a company pioneering artificial intelligence technologies for the discovery and development of small molecule medicines.^[1]

Early life and education

In 2013, Feinberg earned a B.S. in Applied Physics from Yale University.^[2] In 2018, Feinberg received a Ph.D. in computational Biophysics from Stanford University under his doctoral advisor Vijay Pande;^[3] his Ph.D. thesis was entitled “Artificial Intelligence Methods for Molecular Property Prediction.”

While at Stanford, Feinberg co-invented PotentialNet, a machine learning framework designed for molecular property prediction in drug discovery.^[4] PotentialNet uses graph neural networks to model interactions such as protein-ligand binding affinity.^[5] The approach was evaluated for potency prediction and its results were further examined through a collaboration between Stanford and Merck & Co., where Feinberg worked as a deep learning consultant prior to the launch of Genesis Molecular AI.^[1]

Career

Feinberg founded Genesis Molecular AI in 2019 with the goal of leveraging his academic work in AI-driven drug design to accelerate novel small-molecule discovery. He has led the development of the company’s AI-powered platform, GEMS (Genesis Exploration of Molecular Space), which integrates generative and predictive AI with physics for designing small molecule medicines.^[6]

Under Feinberg's leadership, Genesis has raised over $300 million, including a $200 million Series B round in 2023.^[1] Genesis has has also secured notable partnerships with major pharmaceutical and technology companies, including Gilead Sciences, Incyte, and NVIDIA.^[7] In partnership with NVIDIA in 2025, Genesis announced Pearl (Placing Every Atom in the Right Location)^[8], a foundation model for predicting the 3-D structure of protein-ligand complexes.

Awards

In 2019, Feinberg was named a Forbes 30 under 30 recipient for his work at Genesis and creating a neural network that can more quickly and effectively identify potential drug candidates.^[9] He is also a recipient of the Yale Applied Physics Prize^[10] and the Blue Waters Graduate Fellowship.^[5] In 2025, he was named to BioSpace's 40 Under 40 list.^[11]

Publications

Feinberg’s research has appeared in journals, including Science, Nature, ACS Central Science, and the Journal of Medicinal Chemistry.^{[5][8][12][13][14][15]} His notable publications include:

• Genesis Research Team, Dobles, A., Jovic, N., Leidal, K., et al., including Feinberg, E. N. (2025). "Pearl: A foundation model for placing every atom in the right location." arXiv, October 2025. Pearl is a protein-ligand cofolding foundation model that surpasses AlphaFold 3 and other open-source baselines on the Runs N' Poses and PoseBusters benchmarks, delivering improvements of 14.5% and 14.2%, respectively. Key innovations include large-scale physics-based synthetic data training, an SO(3)-equivariant diffusion architecture, and controllable inference. The work also provides the first evidence of synthetic data scaling laws in molecular AI.
• Feinberg, E. N., Sur, D., Wu, Z., et al. (2018). "PotentialNet for molecular property prediction." ACS Central Science, 4(11), 1520–1530. This paper introduced a novel graph convolutional neural network (GNN) framework—PotentialNet—for predicting molecular properties such as protein-ligand binding affinities. The method demonstrated superior performance in pharmaceutical applications and was adopted by companies including Merck.
• Feinberg, E. N., Tseng, H., Husic, B. E., et al (2020). "Improved ADMET property prediction with graph convolutional networks." Journal of Medicinal Chemistry, 63(16), 8835–8848. Building on PotentialNet, this work achieved what the authors described as "unprecedented accuracy" in predicting ADMET (absorption, distribution, metabolism, elimination, and toxicity) characteristics, outperforming traditional machine learning approaches such as Random Forest models.
• JBurg, J. S., Ingram, J. R., Venkatakrishnan, A. J., …  Feinberg, E. N., et al. “Structural basis for chemokine recognition and activation of a viral G protein–coupled receptor.” Science, 347, 1113-1117(2015).
• A. B. Farimani, E. N. Feinberg, V. S. Pande. "Binding Pathway of Opiates to μ-Opioid Receptors Revealed by Unsupervised Machine Learning." arXiv, April 2018.

References