Principal Data Scientist - Agentic AI Discovery & Prototyping

Roche•South San Francisco, CA

4d•Onsite

About The Position

The Principal Data Scientist - Agentic AI Discovery & Prototyping will play a foundational role in building and advancing Genentech’s AI Innovation capabilities within the AI Management, Governance, and Research (MGR) team. This role is responsible for exploring frontier AI capabilities, rapidly translating research breakthroughs into experimental prototypes, and systematically advancing validated innovations into the enterprise AI product ecosystem. Operating at the intersection of AI research discovery, advanced experimentation, and applied product development, this role will explore emerging techniques across Generative AI, predictive modeling, and causal inference, with particular emphasis on large language models, multi-agentic human-in-the-loop systems, and advanced AI tooling ecosystems. The role will focus on identifying high-potential innovations through structured opportunity sensing, designing innovation sprint cycles, and building experimental AI systems that test new architectures, models, and frameworks. This includes evaluating emerging approaches such as agentic orchestration, model fine-tuning, tracing-auditing, alignment strategies, AI measurement strategies, multimodal AI workflows, and advanced model evaluation frameworks. Through a formal R&D operating model, the Principal Data Scientist will lead experimentation initiatives defined by structured research briefs outlining hypotheses, system architectures, evaluation methodologies, and scalability pathways. Successful prototypes will inform and accelerate downstream product development in partnership with AI Product Management, Data Science, and ML Engineering teams. In parallel, this role contributes to shaping the organization’s AI research and innovation ecosystem, including producing internal AI Signals Briefs, evaluating emerging technologies, and identifying opportunities for technical publications, patents, and reusable AI capabilities that strengthen Genentech’s long-term AI leadership. This role requires a strong blend of technical depth, hands-on prototyping ability, and systems thinking, enabling Genentech to responsibly explore emerging AI capabilities while maintaining strong alignment with enterprise priorities, governance frameworks, and real-world product impact.

Requirements

Bachelor’s degree with 10+ years of experience in Data Science, Machine Learning, Artificial Intelligence, or a related field, or Master’s/PhD with 7+ years of experience.
Demonstrated experience working in AI research, advanced experimentation, or applied ML innovation environments.
Deep hands-on experience developing Generative AI systems, including techniques such as: Retrieval-augmented generation (RAG), Prompt engineering and prompt optimization, Model fine-tuning and alignment, Agentic AI systems or multi-agent workflows
Experience working with modern LLM ecosystems and orchestration frameworks.
Thought leadership in Responsible AI principles
Strong foundation in Machine Learning and Predictive Modeling techniques.
Experience designing experiments that evaluate model behavior and AI system performance.
Demonstrated ability to translate emerging research into working prototypes and experimental systems.
Experience building AI applications using modern development frameworks and APIs.
Strong programming skills in Python and familiarity with modern AI frameworks such as: PyTorch, TensorFlow, Hugging Face, LangChain / LlamaIndex / LangGraph or similar
Experience working with vector databases, embedding models, and LLM application architectures.
Demonstrated experience using AI assisted coding tools and IDEs such as Cursor, Claude Code, or Codex.
Experience designing model evaluation frameworks, benchmarking pipelines, or experimentation harnesses.
Understanding of system-level tradeoffs including latency, reliability, cost efficiency, safety, and scalability.
Strong written and verbal communication skills with the ability to translate complex AI concepts into clear technical guidance for diverse audiences.

Nice To Haves

Working knowledge of AI protocols like MCP and A2A in the context of multi-agent orchestration systems.
Experience working with synthetic data generation, model evaluation datasets, or automated testing frameworks for AI systems.
Track record of contributing to research publications, patents, or open-source AI projects.
Experience working within innovation labs, applied research teams, or emerging technology R&D environments.
Experience in healthcare, life sciences, or other highly regulated industries.