Why Achira

At Achira, we are building a team of world-class scientists, ML researchers, and engineers to move beyond the beaten path and build a frontier lab for Physical AI for molecules. We are actively exploring the next frontier of model architectures for AI x Chemistry: developing world models for the physical microcosm. Our goal is to make biology at the molecular level something that can be learned, predicted, and designed.

At Achira, you’ll operate at the frontier scale of massive compute, massive data, and massive ambition. You’ll own impactful work end-to-end, from ideation to architecture to deployment on distributed infrastructure. We are a well-funded, talent-dense organization that values rigor, speed, execution, and an ownership mindset. We’re looking for new members who share our sense of relentless urgency and are natural collaborators who value team success.

About the Role

We’re looking for deep learning researchers who want to build rich representations of atomistic systems to help teach AI about the microscopic world. You will collaborate with domain experts in molecular modeling to develop Achira’s next-generation foundation models. You’ll work at the intersection of model architecture, data, and training strategy to find new points on the Pareto frontier of representational richness, robustness, accuracy, and speed for microscale world models.

While we prefer candidates willing to work from our San Francisco office, highly skilled candidates may be considered for working from New York City with travel to San Francisco as needed. Both locations are offered as hybrid roles, spending at least some of your time working from the office in collaboration with coworkers. Travel is part of all roles at Achira, both to conferences and corporate on-site activities.

What You’ll Do

• Build a robust pre-, mid-, and post-training curriculum that ensures foundation model performance and impact.

• Create reinforcement learning strategies to help models focus their capacity where it matters most, especially when the training data doesn’t cover the domain of applicability.

• Develop expressive representations of molecular and atomistic structure and dynamics, including equivariant graph neural networks, geometric transformers, and latent encoders that capture physical symmetries and constraints.

• Prototype, benchmark, and iterate rapidly to transform research ideas into reusable and scalable components across Achira’s ecosystem.

• Collaborate with physicists and chemists to ensure models are grounded in real physics.

• Work with research engineers and the infrastructure team to identify where research ideas will need support in order to deliver effective results.