LLM-based Agents Intern

2w2 weeks ago

Genentech

New York City, US · Internship · $90,000 – $110,000

About this role

Prescient Design, part of Genentech’s gRED, advances drug discovery through cutting-edge machine learning. The Foundation Models team builds large language models and agent platforms for next-generation scientific and biomedical applications across the drug-discovery pipeline. Join exceptional efforts on one of the largest scientific agent platforms in the industry.

The team develops and operates a foundational platform transforming scientific knowledge and actions from thousands of world-class scientists into sharable tools, workflows, and agents. This reshapes drug discovery with large-scale in-house use cases. The internship is on-site in New York City.

Contribute to bringing LLMs and AI into real-world drug discovery use cases. Collaborate closely with AI/ML engineers and scientists in a fast-paced environment. Interact with world-class drug discovery experts across Roche at various pipeline stages.

Choose engineering track to own end-to-end solutions, develop agent features like reasoning layers, and train LLMs with reinforcement learning. Or select research track to own topics like evaluation methods for agent outputs and publish at top venues. Explore the frontier of LLMs and agents tackling unsolved scientific problems.

Requirements

Must be pursuing a PhD (enrolled student)
Strong background in machine learning and large language models
Experience with AI agents and reinforcement learning techniques
Proficiency in Python and ML frameworks for scientific applications
Familiarity with drug discovery or biomedical data workflows
Ability to tackle unsolved problems in AI for science

Responsibilities

Contribute to cutting-edge efforts bringing LLMs and AI capabilities into real-world drug discovery use cases
Collaborate closely with AI/ML engineers and scientists on engineering and research
Interact with world-class drug discovery experts across Roche
Own end-to-end solution: understand scientists’ needs, build agents and tools, deploy and iteratively improve
Contribute to product design and develop new features for the agent platform
Develop agents capable of building other tools/agents by mining scientific knowledge and workflows
Train underlying LLMs using techniques like reinforcement learning
Own research topic on GenAI usage in scientific scenarios and validate hypotheses