ML Research Scientist - Atomistic Foundation Models

Invent the next generation of deep generative, representational, and simulation models for molecules and materials — building the foundation models that make the atomistic world learnable, predictable, and designable. WHY ACHIRA - Join a world-class, interdisciplinary team of ML researchers, physicists, chemists, and engineers reimagining atomistic simulation through large-scale foundation models. - Push the frontier where deep learning meets the laws of nature — bridging generative AI, probabilistic reasoning, and molecular physics. - Work at a scale few attempt: massive data, massive compute, and massive ambition. - Own your research end-to-end — from concept and architecture to training, evaluation, and deployment. - Thrive in a culture that rewards rigor, velocity, creativity, and impact — not bureaucracy. ABOUT THE ROLE Achira is building foundation simulation models — large-scale models that learn the structure, dynamics, and energetics of the atomistic world. These models unify deep representation learning, generative modeling, and advanced simulation and sampling. As a Generative AI Researcher, you will: - Design and train frontier deep generative models — diffusion, autoregressive, flow-based, and latent-variable architectures — for molecules, materials, and atomic systems. - 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. - Invent advanced sampling and simulation methods that integrate probabilistic inference, deep learning, and reinforcement learning — enabling efficient exploration and simulation of learned energy landscapes. - Build models that understand, generate, and simulate the physical world — unifying reasoning, simulation, and prediction. - Collaborate with physicists and chemists to ground models in ab initio, molecular dynamics, and experimental data. - Prototype, benchmark, and iterate rapidly — transforming research ideas into reusable, scalable model components across Achira’s foundation model stack. - Contribute to publications, open-source tools, and internal research projects that advance the field. ABOUT YOU You are a deep learning researcher who moves seamlessly between representation learning, generation, and simulation — motivated by the idea of teaching AI to reason about the physical world. Required Qualifications - PhD or equivalent research experience in machine learning, physics, chemistry, computer science, or a related field. - Proven expertise in deep generative modeling (e.g., diffusion, VAEs, flows, autoregressive transformers). - Experience in representation learning for structured data, especially graph or 3D geometric models (GNNs, SE(3)/E(3)-equivariant networks, geometric transformers). - Proficiency in Python and modern ML frameworks (PyTorch, JAX) plus scientific libraries (NumPy, SciPy). - Solid grounding in probability, optimization, and deep learning fundamentals. - Demonstrated research impact through publications, open-source contributions, or released models. PREFERRED QUALIFICATIONS - Experience with atomistic simulations, molecular dynamics, or electronic-structure data. - Familiarity with probabilistic inference, MCMC, variational methods, or reinforcement learning for sampling and control. - Experience integrating physics-informed priors or energy-based models into deep architectures. - Knowledge of atomistic molecular datasets and benchmarks such as SPICE, OMol25, QCML, AIMNet2 - Experience scaling models on HPC or distributed GPU infrastructure. - Strong technical communication across interdisciplinary teams. WHAT SUCCESS LOOKS LIKE - You develop models that both represent and generate molecular systems, and simulate their dynamics through learned sampling and reasoning. - Your architectures and algorithms become core components of Achira’s foundation model platform. - You thrive in collabora