Anthropic· Technical Program Management · San Francisco, CA | New York City, NY
Technical Program Manager, Discovery
Classified Tasks (27)
Automate 4%Augment 44%Human-Only 52%
Automate (1)
Fully handled by AI agents
Monitor failure rates and stability of scientific RL environments
technical
Augment (12)
AI assists, human decides
Manage Discovery's compute planning across supervised learning, reinforcement learning, and sandboxing workloads
operational
Forecast compute needs for Discovery workloads
analytical
Allocate compute resources to research and experimental workloads
operational
Prioritize compute workloads and scheduling across teams
operational
Implement efficiency improvements in compute usage and planning
technical
Monitor the quality of scientific RL environments
technical
Monitor reward integrity in scientific RL environments
analytical
Conduct quality reviews of vendor-supplied RL environments
analytical
Design reward functions and reward structure for RL environments
technical
Integrate vendor RL environments into production pipelines
technical
Debug data pipelines used for research and experimentation
technical
Read RL transcripts and logs to spot issues and anomalies
analytical
Human-Only (14)
Requires human judgment
Own systems and programs that determine research velocity, including compute planning, scientific RL environment health, and vendor pipelines
leadership
Partner with central compute planning to represent and secure Discovery's compute needs
communication
Drive issues in RL environments to resolution
operational
Expedite the external vendor pipeline for RL environments
operational
Work with research teams across life sciences, STEM, and other domains to define and translate research goals
communication
Translate research goals into roadmaps that advance AI scientist capabilities
leadership
Establish processes and frameworks to bring structure to an unstructured research setting
operational
Collaborate with research leads, infrastructure engineers, and data operations to identify blockers
communication
Prioritize competing technical and operational needs across teams
leadership
Make technical trade-off decisions among research, infrastructure, and operational constraints
leadership
Incubate new programs in domains such as bio R&D
creative
Make allocation and quality decisions in real time when experimental or production runs hit problems
operational
Identify critical people and teams across research, infrastructure, product, and data operations
communication
Coordinate cross-functional teams without losing project velocity
leadership
Job description
About Anthropic 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. About the Team The Discovery team is organized around the north star of building an AI scientist — a system capable of solving the long-horizon reasoning challenges and core capabilities needed to push the scientific frontier. The team trains large-scale models, runs complex multi-week experiments, and builds novel products at the intersection of AI and science. About the Role As a Technical Program Manager on the Discovery team, you will own the systems and programs that determine how fast our research moves: compute planning, scientific RL environment health, and the vendor pipelines that supply them, with scope to incubate new programs in domains like bio R&D. Strong candidates should have an ML engineering or research background and have grown into program leadership. You'll need real technical depth: the ability to debug data pipelines, read RL transcripts to spot issues, and make allocation and quality decisions in real time when experimental or production runs hit problems. You'll need organizational effectiveness in equal measure: the ability to navigate a fast-growing organization, quickly identify the critical people and teams across research, infrastructure, product, and data operations, and coordinate across them without losing velocity. Join us in our mission to build AI systems that push the frontiers of science and benefit humanity. Responsibilities Manage Discovery's compute planning across supervised learning (SL), reinforcement learning (RL), and sandboxing workloads, including forecasting, allocation, prioritization, and efficiency improvements. Partner with central compute planning to ensure Discovery's needs are represented and met. Monitor the health of scientific RL environments (quality, reward integrity, failure rates) and drive issues to resolution. Expedite the external vendor pipeline for RL environments, including quality review, reward design, and production integration. Work with research teams across life sciences, STEM, and other scientific domains to translate research goals into roadmaps that advance AI scientist capabilities. Establish processes and frameworks that bring structure to an unstructured research setting without slowing researchers down. Collaborate with research leads, infrastructure engineers, and data operations to identify blockers, prioritize competing needs, and make technical trade-off decisions. You May Be a Good Fit If You: Have a background in ML engineering, ML research, or STEM R&D before transitioning to technical program management Have deep, hands-on experience with ML training pipelines, RLHF systems, and large-scale data infrastructure in production. Have a track record of building execution plans and inventing high-leverage processes that reduce operational overhead and let researchers focus on research. Are a fast learner who builds deep contextual understanding in unfamiliar technical domains and can contribute meaningfully to discussions with researchers. Are resourceful, high-agency, and able to navigate ambiguity and shifting priorities to drive progress in a fast-moving research setting Have excellent stakeholder management and communication skills, with the ability to influence senior technical staff through clarity, competence, and consistent delivery. Are excited about the potential for AI to accelera