ML Engineer

We offer opportunities to do your life’s work while helping solve one of the most important technical and moral challenges of our time. Full-time, in-office in Emeryville, California. Compensation includes equity. Build the reliable systems that carry our data from an assay recording to a reproducible model, an evaluated prediction, and a usable recommendation for the next experiment. Key Responsibilities • Own pipelines for ingesting, validating, versioning, and joining assay videos, metadata, compound records, model features, and experimental outcomes • Build reproducible training and evaluation infrastructure with clear data lineage, model versioning, automated tests, and auditable outputs • Turn research prototypes into dependable batch and online systems that can rank compounds and surface recommendations through our tools • Monitor data quality, distribution shift, calibration, latency, cost, and failures as the number of labs and assays grows • Design interfaces between computer vision, molecular models, active-learning systems, and the lab workflow • Improve developer and researcher velocity without weakening scientific reproducibility or access controls Qualifications • Strong production software engineering experience in Python and modern machine-learning or data systems • Experience deploying and operating model-training, feature, evaluation, or inference pipelines in a cloud environment • Fluency with testing, observability, data validation, version control, and reproducible computational workflows • Ability to work with large video datasets and structured scientific data • Ability to collaborate closely with researchers while making sound engineering tradeoffs Desired Attributes • Experience with PyTorch, JAX, or TensorFlow and workflow-orchestration tools • Experience on Google Cloud or with large-scale object-storage pipelines • Familiarity with computer vision, molecular machine learning, active learning, or scientific data platforms • Instinct for simple systems, explicit failure modes, and measurable reliability