Applied AI Engineer, Life Sciences; Beneficial Deployments San Francisco, CA NY

Overview

Applied AI Engineer, Life Sciences (Beneficial Deployments)

Anthropic’s mission is to create reliable, interpretable, and steerable AI systems. We want AI to be safe and beneficial for our users and for society as a whole. Our team is a quickly growing group of committed researchers, engineers, policy experts, and business leaders working together to build beneficial AI systems.

Beneficial Deployments ensures AI reaches and benefits the communities that need it most. We partner with nonprofits, foundations, and mission-driven organizations to deploy Claude in education, global health, economic mobility, and life sciences — focusing on raising the floor for those who need it most.

We re looking for an Applied AI Engineer to join our Beneficial Deployments team, focused on maximizing the impact of Claude in the life sciences. Our goal is ambitious: accelerate scientific progress from R&D through translation by an order of magnitude. That means making Claude the go-to tool for the life sciences ecosystem from early discovery in academia to paradigm shifting biotech to reimaging pharma pipelines — and building the technical infrastructure to back that up.

You ll work directly with flagship research partners like Howard Hughes Medical Institute and The Allen Institute, embedded in their scientific workflows. This isn t consulting from the outside — you ll be building alongside their engineers, prototyping agents that fit into real research pipelines, and developing the ecosystem-level tooling (MCP servers, benchmarks, reusable agent skills) that extends Claude s usefulness across the broader life sciences community.

This role will be part of the founding Beneficial Deployments applied AI team focused on bringing more of life sciences closer to the frontier and be responsible for building with our partners.

• Work as a deep technical partner to flagship life sciences research institutions — understanding their scientific workflows end to end, then advising on where AI can meaningfully accelerate discovery. This means conversations with both PIs and engineering teams, and the ability to hold your own in both.
• Build hands-on with partner engineering teams. Pair program, prototype, contribute code.
• Help a research team go from "we think Claude could help with our analysis pipeline" to a production system that s actually integrated into how they do science.
• Develop the ecosystem infrastructure that makes Claude useful across life sciences broadly — MCP servers connecting Claude to domain-specific data sources (genomics platforms, literature databases, experimental repositories), instruments, benchmarks grounded in real scientific tasks, and reusable agent skills that other institutions can adopt without starting from scratch.
• Help design and evaluate agentic scientific workflows — the kind of systems where Claude isn t just answering questions but actively contributing to experimental design, analysis, and interpretation.
• Identify what s actually hard about deploying AI in life sciences — the heterogeneous data, the need for auditability, the gap between a promising prototype and something a researcher will trust — and feed those findings back to product, engineering, and research to make the platform better.
• Create technical content and documentation that lets life sciences partners self-serve. If you build something that works at for one partner, we want it to work for institutes and scientists around the world, without requiring the same level of hand-holding.

• 4+ years as a Software Engineer, Forward Deployed Engineer, or technical founder — with production experience shipping systems that real users depend on.
• Deep experience building LLM-powered applications: prompting, context engineering, agent architectures, evaluation frameworks, deployment  ve debugged evals at 2am and you know what it takes to get from "works in the notebook" to "works in production."
• Meaningful exposure to life sciences, biomedical research, or scientific computing. Enough to understand why a researcher might be skeptical…