Machine Learning Research Engineer, Agent Data Foundation - Enterprise GenAI

Scale AI, Inc.•San Francisco, CA

51d

About The Position

AI is becoming vitally important in every function of our society. At Scale, our mission is to accelerate the development of AI applications. For 9 years, Scale has been the leading AI data foundry, helping fuel the most exciting advancements in AI, including generative AI, defense applications, and autonomous vehicles. With our recent investment from Meta, we are doubling down on building out state of the art post-training algorithms to reach the performance necessary for complex agents in enterprises around the world. The Enterprise ML Research Lab works on the front lines of this AI revolution. We are working on an arsenal of proprietary research, tools, and resources that serve all of our enterprise clients. As MLRE on the Data Foundation team, you'll work on cutting edge research to define the data flywheel that makes the whole machine move. This includes research around synthetic environments from task definitions, building agents for trace analysis, and contributing to a cutting edge framework that automatically hill-climbs agent-building from an eval set. This will involve creating best-in-class Agents that achieve state of the art results through a combination of post-training + agent-building algorithms. If you are excited about shaping the future of the modern GenAI movement, we would love to hear from you!

Requirements

3+ years of building with LLMs in a production environment
Clear experiences with constructing high quality data to use to improve an LLM/Agent
Publications in top conferences such as NEURIPS, ICLR, or ICML within the last two years
PhD or Masters in Computer Science or a related field

Responsibilities

Build synthetic data pipelines to generate enterprise environments to use for RL post-training
Create agents to convert traces from production into actionable insights to use to improve agents
Contribute to our agent building product which can construct other agents using coding agents + proprietary algorithms
Train state of the art models, developed both internally and from the community, to deploy to our enterprise customers.