Senior Machine Learning Research Engineer

About Scale

Scale’s mission is to develop reliable AI systems for the world’s most important decisions. As the leading AI data foundry, we provide the high-quality data and full-stack technologies that power the world’s most advanced models — fueling breakthroughs in generative AI, defense, and autonomous vehicles. We partner with leading enterprises and governments to bring AI into production that performs when it matters most, combining rigorous evaluation with full-stack deployment so our customers can build AI they can trust.

Applied Intelligence Systems (AIS) is part of the Scale Generative AI Platform (SGP), focused on pushing the frontier of what agentic applications can do across diverse enterprise and government use cases. We build the infrastructure and tooling that power agentic AI in production, paired with applied ML research, design, and evaluation to ensure these systems perform reliably at the scale our customers demand.

AIS spans multiple work streams — agent evaluation and oversight, orchestration and tool-use infrastructure, model and systems optimization, and applied research on new agent capabilities — and this role is not scoped to any single one of them. We’re growing fast, with increasing traction across both commercial and public sector customers, and we’re just getting started — this team will define what dependable, production-grade agentic AI looks like.

As a Staff Machine Learning Research Engineer, you will operate across the full breadth of AIS’s technical needs — wherever the hardest ML problem in agentic AI happens to be that quarter. This could mean training and fine-tuning models, designing evaluation and observability systems, building improvement loops from production data, prototyping novel agent architectures, or designing internal systems and tooling that boost productivity across teams.

You’re not tied to one team’s roadmap; you’re expected to move to where the technical leverage is highest, and to set the AI/ML technical direction across AIS — the methods, architectures, and standards other teams build on, not just your own workstream. This is a hands-on research and engineering role at staff scope: you’ll write code — training pipelines, evaluation systems, infrastructure, or whatever the problem calls for — and ship production systems yourself, while also setting AIML technical direction and raising the bar for engineers and scientists across AIS.

• Move across AIS’s core problem areas as needed — training/fine-tuning, inference, memory and retrieval, evaluation and observability, orchestration and tool-use infrastructure, applied research on new agent capabilities — going wherever the technical leverage is highest rather than owning one fixed surface

• Research and prototype novel methods for agent performance improvement in a production/enterprise-ready setting — continuous learning loops, automated curriculum or data generation from production traces, online or offline RL — and validate them with rigorous experiments before they ship, making the call on where to build new infrastructure versus apply existing methods

• Build AI agents and internal tooling that reduce bottlenecks in AIS’s own processes — cutting down time spent on repetitive evaluation, data, or experimentation work so teams can focus on the hard problems

• Partner with other ML engineers, software engineers, product managers, customers, data annotators, and Forward Deployed Engineers to take your work from idea to production and translate enterprise and government requirements into robust ML capabilities

• Set AI/ML technical direction, mentor senior and staff-track engineers and scientists across teams, and raise the bar on experimental rigor org-wide

• 5+ years of experience as an ML engineer or applied/research scientist, including direct experience training or fine-tuning models in production systems

• PhD in Computer Science, Electrical Engineering, or a related field

• Broad, hands-on fluency across the agentic ML stack — model training and fine-tuning (SFT, RLHF/RLAIF, reward modeling), evaluation and observability infrastructure, and agent…