Principal ML Engineer: Foundation Models for Drug Discovery

Roche; South San Francisco (CA); On-site; USD 193,000 - 358,000; Full time

Job summary

Roche Genentech’s AI for Drug Discovery (AIDD) group is hiring a Principal Machine Learning Engineer to scale foundation models and agentic systems for autonomous drug discovery workflows. You will work across data, infrastructure, and MLOps to enable robust scientific computing at scale.

You will own end-to-end ML infrastructure, lead technical strategy, and partner with ML scientists to translate open-ended problems into production-grade systems.

Qualifications

• BS/MS/PhD in CS/ML with leadership in ML infra and end-to-end ownership.
• Strong Python programming and software engineering fundamentals.
• Hands-on experience with PyTorch or JAX and distributed training.
• Experience with agent orchestration frameworks and tool integration.

Responsibilities

• Agentic Systems Engineering: design autonomous agents with memory for long-horizon tasks.
• Scalable ML Systems: build distributed training and inference for foundation models.
• MLOps/AgentOps: establish CI/CD, observability, evaluation harnesses and infra.
• Technical Strategy: define roadmaps and elevate ML infra across teams.
• Research-to-Production: translate scientific problems into producible systems.

Skills

Python

Software engineering fundamentals

Git

CI/CD

PyTorch

JAX

AWS

LangGraph

MCP-based tool integration

Agent orchestration

Distributed systems

Education

PhD in Computer Science / ML

MS in Computer Science / ML

BS in Computer Science / Engineering

Tools

LangGraph

MCP

Python tooling

Job description

Roche Genentech’s AI for Drug Discovery (AIDD) group is hiring a Principal Machine Learning Engineer to scale foundation models and agentic systems for autonomous drug discovery workflows. You will work across data, infrastructure, and MLOps to enable robust scientific computing at scale.

You will own end-to-end ML infrastructure, lead technical strategy, and partner with ML scientists to translate open-ended problems into production-grade systems.