ML Research Scientist - Atomistic Simulation Models

Invent and exploit probabilistic generative models that exploit to Achira’s foundation simulation models for drug discovery to accelerate generative molecular design and biomolecular conformational sampling. 🚀 WHY ACHIRA - Join a world-class team of researchers, scientists, and engineers unifying probabilistic AI/ML and molecular simulation to reimagine small molecule drug discovery. - Advance new architectures for conditional 3D generation and learned proposal mechanisms informed by physical priors. - Operate at the frontier scale of large models, large datasets, and high-throughput evaluation on an ML-framework–native biomolecular simulation stack. - Own impact end-to-end from model conception to sampler design to prospective design tools. - Work in a culture that rewards rigor, speed, and scientific depth with an ownership mindset. ABOUT THE ROLE Achira is building foundation simulation models http://achira.ai and conditional generators for molecular systems. You will design probabilistic generative models (utilizing strategies such as diffusion models, normalizing flows, and flow matching) that that exploit Achira’s next-generation biomolecular simulation potentials. Your work will enable target- and property-conditioned small-molecule generation and efficient exploration of biomolecular conformational landscapes, driving measurable gains in efficiency for small molecule design. Familiarity with statistical mechanics—particularly nonequilibrium statistical mechanics based on Crooks/Jarzynski viewpoints—is desirable, but the center of gravity is probabilistic AI/ML. 🛠️ WHAT YOU’LL DO - Develop conditional molecular generators: Build conditional small-molecule generators (e.g., pocket/scaffold/pharmacophore- and property-conditioned) using generative modeling strategies such as diffusion models, normalizing flows, and flow matching with 3D- and symmetry-aware representations. - Develop efficient samplers: Develop sequential sampling pipelines (e.g. SMC/AIS/tempering/Boltzmann generators) that anneal from learned priors into probabilities induced by Achira’s ML potentials, maximizing ESS and reducing bias/variance. - Couple learning and sampling: Design learned proposal mechanisms (transport maps, score-guided moves) that adapt to stiff, multimodal landscapes and improve mixing and wall-clock efficiency. - Leverage nonequilibrium statistical mechanics: Where beneficial, use nonequilibrium switching protocols and work-based estimators to accelerate exploration and estimate partition-function ratios/affinity proxies. - Measure what matters: Define and track relevant metrics (ESS/compute, acceptance probabilities) and build reliable evaluation harnesses for fast, physics-informed feedback. - Experiment and engineer for reproducibility: Collaborate with our engineering team to implement robust research software in Python (PyTorch and/or JAX), with tests, CI, experiment tracking, and clear documentation. - Collaborate closely: Partner with computational chemistry, AI/ML, and platform teams to shape objectives (potency, selectivity, developability) and run prospective design studies. - Automate workflows: Use generative coding and experiment-management tools to accelerate iteration and close active-learning loops with synthetic data generation in the loop. 🧠 ABOUT YOU - Probabilistic ML background: Deep grasp of probabilistic machine learning, Markov chain Monte Carlo, variational inference, diffusion models, normalizing flows, flow matching, and uncertainty quantification. - Sequential methods expert: Experience with sequential Monte Carlo methods, proposal design, and diagnostics for high-dimensional, multimodal targets. - Geometric intuition: Comfort with graph/point-cloud/SE(3)-aware models and constraints relevant to protein–ligand systems and conformer generation. - Systems thinker: You integrate models into end-to-end pipelines (data → model → sampler → physics-aware evaluation → candidate triage) and care about measurable impact. - Familiarity with statistical mechanics(nice to have): Working knowledge of statistical mechanics, sampling, estimators, and the Crooks/Jarzynski perspective of nonequilibrium statistical mechanics will be a superpower. - Engineering discipline: Strong Python skills with PyTorch and/or JAX, Git/CI/testing, and reproducible experiment management. - Mindset: You value rigor, move with urgency, collaborate well, and enjoy turning ideas into reliable, high-impact tools. ✅ MINIMUM QUALIFICATIONS - PhD (or equivalent research experience) in computer science, statistics, applied math, computational chemistry/biology, or related field. - Demonstrated track record in probabilistic ML and generative modeling (publications, impactful open-source, or deployed systems). - Industry experience tackling practical machine learning problems. - Hands-on experience with diffusion/flows/flow matching on structured or geometric data. - Practical experience with sequential Monte Carlo/AIS/tempering and/or advanced MCMC. - Proficiency in Python with PyTorch and/or JAX; strong software engineering hygiene. - Familiarity with biomolecular structure and data representations (graphs/3D/SMILES). ⭐ PREFERRED QUALIFICATIONS - Experience with ML interatomic/energy potentials is a bonus - Background in SE(3)-equivariant architectures, geometric deep learning, or score matching on manifolds. - Experience with active learning / Bayesian optimization or RL-style acquisition for proposal selection. - Experience with implementing MCMC sampling approaches grounded in statistical mechanics—especially nonequilibrium approaches that utilize Crooks/Jarzynski—a plus. - Contributions to open-source scientific software; experience mentoring or leading small research efforts. ELIGIBILITY In compliance with United States federal law, all persons hired will be required to verify identity and eligibility to work in the United States and to provide required employment eligibility verification documentation upon hire.