Description
This is a deliberately broad ML role for someone who does not want to spend several years working on one narrow component of the agent stack.
Scale’s Applied Intelligence Systems group builds production-grade agentic AI for enterprise and government applications. The Staff/Senior ML Research Engineer moves between whichever technical problems currently have the highest leverage: training and fine-tuning, inference, memory, retrieval, evaluation, observability, tool-use infrastructure, or new agent architectures.
One project might involve improving a continuous-learning loop from production traces. Another could require building an automated curriculum generator, developing an RL intervention, designing evaluation infrastructure, or prototyping a new way for agents to plan and use tools.
The important distinction is that research is expected to reach production. Engineers are not only asked to produce experimental results; they are expected to write production code, build the infrastructure required to validate an approach, and work with software engineers to deploy successful methods.
The agent stack includes SFT, RLHF/RLAIF, reward modeling, evaluation, memory, planning, tool use, orchestration, and potentially multi-agent systems. The strongest candidate should understand several of these areas deeply enough to switch contexts when the organization’s priorities change.
Collaboration also extends beyond research. Product managers, Forward Deployed Engineers, customers, data annotators, ML engineers, and software engineers can all participate in turning an experimental technique into something that works under real enterprise constraints.
Scale is seeking substantial ML experience and a research-oriented background, with a PhD listed among the requirements. Staff-level candidates should also be able to set technical direction and influence methods or architectures adopted by other teams.
Published work, continuous fine-tuning, online learning, agent training, optimization, and regulated-environment experience are all useful additions.
The appeal of this role is breadth: rather than optimizing one benchmark in isolation, the engineer helps define what reliable production agentic AI should look like across several layers of the system.
Curated opportunity. Please verify details and apply via the original link below. No Signals are required for this project/job.
https://scale.com/careers/4714527005