Machine Learning Scientist, Generative Molecular Design; Material Discovery

Machine Learning Scientist, Generative Molecular Design (Material Discovery) at SandboxAQ, United States (remote). A mid-level full time position in the Chem Sim team, listed under software engineering.

Introduction to the team:
The AI Simulation (AI Sim) division at SandboxAQ applies physics and AI simulation to problems in drug discovery and the material sciences. You will work within a highly skilled, multidisciplinary team of computational chemists, physicists, AI experts, and software engineers, on a generative platform that is used across both application areas.

Introduction to the role:
We are looking for a Machine Learning Scientist to help build the generative engine at the centre of SandboxAQ's material discovery work. The platform finds compounds that match a desired property profile by navigating chemical space while balancing multiple competing objectives. You will: (1) implement, extend, and optimize frontier generative architectures for material discovery; (2) own the interface between the generative engine and the property predictors that supply its reward signal;

(3) identify, ingest, and quality-control the data these models are trained and evaluated on; and (4) turn the models you build into production capability that discovery scientists across the company can run. See how SandboxAQ is helping build America's semiconductor supply chain from the materials up

We are growing this team across three areas of generative expertise, listed in the order we most need them. You do not need all three — deep, hands‑on strength in any one of them is what we are looking for.

• Active learning and bayesian optimization. Driving a design‑make‑test loop with an expensive oracle such as a physics simulation (DFT, molecular dynamics) or an experimental lab‑in‑the‑loop assay.
• Guided generative modeling for material discovery; generation constrained to makeable chemical space; inference‑time steering (guided diffusion, sequential monte‑carlo and related approaches) or fine‑tuning (LoRA, reinforcement learning) for adapting general models to a particular domain of interest;.

• Architect and evolve our flagship generative platform, serving as a core contributor to the engine that will redefine how the industry discovers new materials.
• Implement and optimize state‑of‑the‑art generative models - including diffusion, flow‑matching, and GFlowNets, or LLM- and agent‑based methods — to navigate high‑dimensional chemical space and solve the multi‑objective optimization problems that material discovery presents.
• Own the reward interface. Integrate learned surrogates, and simulation‑derived oracles as the objective signal driving generation, and design reward formulations that hold up when several objectives are in genuine tension e.g. performance against environmental and safety constraints.
• Develop and deploy computational methods and workflows to generate and evaluate hypotheses, guiding design decisions and influencing project direction.
• Identify relevant data sources and take responsibility for data ingestion, QC, and validation.
• Collaborate daily with a core group of ML researchers to implement frontier generative architectures, while working closely with the broader R&D, discovery, engineering, and product teams.

• Hands‑on expertise building and shipping modern generative or applied ML for material discovery with at least 4+ years of experience delivering these solutions in industry or an equivalent research setting.
• Demonstrated ability to write clean, maintainable, and robust production‑grade code designed for industry‑scale applications.
• Proven ability to build scalable ML, e.g. using PyTorch, Tensor Flow, or JAX.
• Familiarity with cloud‑based platforms (e.g. Google Cloud Platform, AWS) for compute and data processing and storage.
• PhD or MSc in Computational Chemistry, Physics, Computer Science, or equivalent.
• Excellent communication skills and a record of fostering successful internal collaboration in a fast‑paced environment.
• Experience developing ML code in a collaborative environment using modern software engineering best practices.