The Opportunity
At Genentech (gRED) and Roche (pRED), we’re applying software, robotics, and advanced AI to solve one of humanity’s hardest challenges: building fully autonomous, self-healing laboratories for lights-out drug discovery. Advances in AI, data, and computational sciences are transforming drug discovery and development. The Computational Sciences Center of Excellence (CoE) and the Data and Digital Catalyst (DDC) team are building the foundational platform for autonomous labs—turning frontier AI models, rich physical telemetry, and multi-vendor robotics into resilient, intelligent systems that run complex science unattended around the clock to deliver life-saving medicines to patients faster than ever before.
As a Senior Multimodal Machine Learning Scientist, you will be the hands-on technical lead and architect developing intelligent systems for autonomous laboratories. Your work will span computer vision and multimodal perception, predictive modeling, agentic systems, and closed-loop experimentation, for example, unifying camera feeds with rich physical telemetry to give lab systems situational awareness and autonomous recovery capabilities. You will build and deploy models and end-to-end ML pipelines, from data collection and model development to evaluation and deployment, helping pioneer the next generation of intelligent laboratory infrastructure at global scale.
In this role, you will:
- Develop and fine-tune state-of-the-art vision models (VLMs, DinoV2, SAM, custom detection/segmentation) to perform real-time visual quality control across lab workcells (e.g., liquid level sensing, tip verification, labware orientation, spill/foam detection).
- Develop multimodal models that integrate camera feeds and video streams with time-series sensor telemetry, system logs, assay metadata, and other laboratory data to catch and understand physical failure modes.
- Build predictive AI models on rich analytical data (spectroscopy, plate-readers, chromatography) to quantify assay quality and support closed-loop experimental decision-making.
- Build the data and ML pipelines needed to support these systems end-to-end, including image/video and sensor data collection, curation, processing, labeling, model training, deployment, monitoring, and iterative improvement.
- Build agentic pipelines that interpret lab data, reason through multi-step experimental workflows, and execute autonomous recovery decisions.
- Build rigorous evaluation suites (evals) to systematically benchmark model and agent accuracy, robustness, and decision safety in drug discovery labs.
- Deploy and optimize scalable ML inference and monitoring infrastructure that is low-latency and can support diverse models, instruments, and scientific use cases.
Who you are:
- MS/PhD in Computer Science, Data Science, AI, or a related quantitative field with 3+ years of hands-on experience deploying production ML systems or translational research models.
- Passion for translating frontier AI into biological workflows and real-world therapeutic breakthroughs.
- Strong background in Computer Vision and Vision-Language Models (VLMs), with hands-on experience training, fine-tuning, or evaluating visual perception architectures (e.g., PyTorch, OpenCV, torchvision, object detection/segmentation, vision-text transformers).
- Strong foundations in machine learning, statistics, and data analysis, with hands-on experience developing and deploying predictive and data-driven models.
- Proven track record building and deploying multimodal models that combine camera feeds/image streams with unstructured physical data (sensor logs, time-series, tabular scientific data).
- Hands-on experience building, orchestrating, or tuning agent harnesses.
- Demonstrated mastery in building comprehensive evaluation pipelines (evals) to benchmark model and agent accuracy, robustness, and safety in production environments.
- Demonstrated ability to work across disciplines and translate scientific requirements into practical ML systems, collaborating with scientists, automation/robotics engineers, and software engineers.
Preferred, but not required
- Experience with synthetic vision data generation, video-based anomaly detection, or edge-vision deployment.
- Familiarity with lab automation hardware, analytical instruments (liquid handlers, plate readers, HPLC), or industrial IoT telemetry.
- Exposure to Bayesian optimization, active learning, reinforcement learning, or Design of Experiments (DoE) for closed-loop experimentation.
- Public GitHub/GitLab portfolio showcasing production ML pipelines, computer vision projects, agentic frameworks/evals, or multimodal models.
Relocation benefits are not available for this opportunity
The expected salary range for this position, based on the primary location of San Francisco, is $168,100 - 312,300. Actual pay will be determined based on experience, qualifications, geographic location, and other job-related factors permitted by law. A discretionary annual bonus may be available based on individual and Company performance. This position also qualifies for the benefits detailed at the link provided below.
#ComputationCoE
#tech4lifeComputationalScience
