Staff ML Engineer - RL & AI Infra for Frontier Models

Fdetimes; San Francisco (CA); Vor Ort; USD 250.000 - 280.000; Vollzeit

Zusammenfassung

Labelbox is seeking an engineer to shape frontier AI by building scalable RL infrastructure and post-training workflows. You will run experiments, create environments agents interact with, and build verifiers to assess success at scale, while delivering reliable, producible platform components.

The role combines systems engineering with agent-focused post-training tasks, demanding strong Python expertise and a pragmatic, high-velocity startup mindset.

Qualifikationen

• 3+ years of experience shipping systems used by customers and engineers.
• High throughput with high quality; ship a lot, review a lot, and establish a solid foundation.
• Strong system and API design judgment requiring decisive architectural calls.
• Production-grade code is expected daily; know how to keep it reliable.
• Build the substrate others rely on — tooling, CI, harnesses, libraries.

Aufgaben

• Design and implement RL environments for agentic tasks including tool surfaces, state/reset semantics, and rewards.
• Develop verifiers and graders, rubric pipelines, and scalable pass@k scoring.
• Create fine-tuning pipelines turning evaluation data into model improvements (SFT & RL).
• Build eval systems to measure model and product quality across millions of agent trajectories.
• Develop training and serving infrastructure to scale workload and cost accounting.

Kenntnisse

Python

System design

API design

Production code

Startup pace

Tools

GraphQL

Kubernetes

GCP

MySQL

PostgreSQL

TypeScript

Jobbeschreibung

Labelbox is seeking an engineer to shape frontier AI by building scalable RL infrastructure and post-training workflows. You will run experiments, create environments agents interact with, and build verifiers to assess success at scale, while delivering reliable, producible platform components.

The role combines systems engineering with agent-focused post-training tasks, demanding strong Python expertise and a pragmatic, high-velocity startup mindset.