Senior Machine Learning Engineer, Public Sector

Excerpt from the original posting

The goal of a Senior Machine Learning Engineer at Scale is to own how we apply generative AI, agentic AI, computer vision, and reinforcement learning to mission-critical problems in production. Our senior machine learning engineers are handed problems that don't yet have an established approach, they propose the architecture, build it with support from other engineers, and are accountable for whether it holds up in the environments our customers depend on.

Our Public Sector Machine Learning team is focused on deploying cutting-edge models to mission-critical government systems through products like  Donovan  and  Thunderforge . Our work spans multiple modalities, with our primary focus on agentic systems built on large language models. We are developing agents that solve complex operational and planning challenges for government partners: agent frameworks that integrate custom retrieval pipelines and production APIs, memory and context-management systems that hold state across long-running tasks, geospatial reasoning over maps and spatial data, and the evaluation tooling that benchmarks and refines agent behavior. We also apply reinforcement learning in targeted places where it earns its keep, and our computer vision work advances evaluation, labeling efficiency, and multimodal model training in support of defense applications.

As a Senior MLE, you'll have design authority over a capability area - the final say on the patterns used within your team, and the responsibility to make those patterns work under real constraints: classified environments, limited compute, and correctness requirements that don't bend.

You will:

- Own the design and delivery of agent capabilities end to end - architecture, implementation, and the evaluation that proves they work

- Define net-new patterns in problem spaces with no established approach, propose them to the wider team, and lead the work to build them

- Take state of the art models developed internally and from the community and put them into production to solve problems for our customers and taskers

- Improve and maintain production models and agents through retraining, hyperparameter tuning, and architectural updates, while preserving core performance characteristics

- Build agent-level evaluation benchmarks, LLM judges, and verifiers - and use it to hillclimb performance rather than just report on it

- Partner with product and research teams to scope and shape high-impact initiatives, includin…

→ Scale AI · Greenhouse