# Numerion Labs

> Numerion Labs (formerly Atomwise) is an AI-native drug discovery company that pioneered the use of deep learning for structure-based drug design. Its AI chemistry superplatform - built on the COSMOS foundation model, the APEX hyper-scalable enumerator, and EXPO optimization algorithms - explores billions of drug-like molecules to find and optimize novel chemical matter. Based in San Francisco with a San Diego presence, the company is advancing its own pipeline of small-molecule programs in immune and inflammatory disease while collaborating with partners across the pharmaceutical industry.

- **Founded:** 2012
- **Headquarters:** San Francisco, California, United States
- **Founders:** Abraham Heifets (Co-Founder (former CEO)), Izhar Wallach (Co-Founder, CTO), Alex Levy (Co-Founder)
- **Team size:** Approximately 17-50 employees
- **Products:** AI Chemistry Superplatform, COSMOS, APEX (Approximate-but-Exhaustive Search), EXPO, Internal Drug Programs
- **Notable:** Invented the use of deep convolutional neural networks for structure-based drug design (as Atomwise, 2012)., Delivered hit-identification success across 230+ academic and collaboration projects., Developed APEX, screening 10 billion compounds in under 30 seconds on a single NVIDIA GPU.

## Products & services

- **AI Chemistry Superplatform** — The company's core drug-hunting engine that explores chemical space to discover and optimize novel, drug-like molecules unseen by others.
- **COSMOS** — A universal chemistry foundation model that predicts drug function from chemical structure, enabling consistent discovery of diverse, novel-structured compounds.
- **APEX (Approximate-but-Exhaustive Search)** — A GPU-accelerated enumerator that pairs deep learning surrogates with structured chemical-space enumeration to screen 10 billion compounds in under 30 seconds on a single NVIDIA GPU.
- **EXPO** — Expert optimization algorithms that deploy project-specific model expertise rapidly without large datasets or retraining, improving conversion of novel binders into clinical candidates.
- **Internal Drug Programs** — A proprietary pipeline of small-molecule therapeutic candidates focused on immune and inflammatory disease areas.

## Achievements

- Invented the use of deep convolutional neural networks for structure-based drug design (as Atomwise, 2012).
- Delivered hit-identification success across 230+ academic and collaboration projects.
- Developed APEX, screening 10 billion compounds in under 30 seconds on a single NVIDIA GPU.
- Published peer-reviewed APEX research with NVIDIA on arXiv with open-source code (October 2025).
- Raised approximately $175M in total venture funding.

## Latest updates

- **2025-10** — Published breakthrough APEX research with NVIDIA on ultra-fast screening of hyper-scalable virtual libraries.
- **2025-02** — Steve Worland, Ph.D. appointed CEO of Numerion Labs.
- **2025-11** — Relaunched brand and website as Numerion Labs (formerly Atomwise), introducing the COSMOS / APEX / EXPO superplatform.

## Links

- Website: https://numerionlabs.ai
- LinkedIn: https://www.linkedin.com/company/numerion-labs
- GitHub: https://github.com/numerionlabs

---

Profile page: https://yespress.io/numerion-labs
Published by YesPress — https://yespress.io
Last updated: 2026-06-13