Jobs at Scale AI

This role involves developing AI applications, focusing on LLMs and agent frameworks, and requires research skills in ML with experience in data strategy.

This role involves building and optimizing ML systems and frameworks, collaborating with teams to enhance research and development in ML models.

Research and develop novel post-training methods such as SFT, RLHF, reward modeling, and preference optimization for text and multimodal LLMs. Analyze model behavior, improve alignment and generalization, mitigate bias, enhance robustness, and publish findings at leading AI conferences. Collaborate with engineers, researchers, and foundation model labs to establish data-driven AI development practices and provide technical and strategic guidance.

Conduct research on evaluating frontier large language models and AI agents. Analyze model behavior, diagnose capability, reasoning, robustness, and alignment failures, and develop benchmarks for text and multimodal systems. Apply post-training techniques such as supervised fine-tuning, RLHF, and reward modeling to connect failures with training interventions. Collaborate with AI labs, define evaluation best practices, translate findings into technical strategy, and publish research at major conferences.

Lead architecture and technical direction for Scale's Generative AI Data Engine, embedding into high-priority Contributor teams to unblock delivery, overhaul platform infrastructure for LLM-native workloads, ensure scalability/reliability, mentor engineers, partner with product/leadership, and own large end-to-end backend, frontend, and ML system initiatives.

The Research Advisor will provide high-level consultancy for generative AI systems, engage in research collaborations, and represent the Human Frontier Collective in key executive meetings.

Perform part-time quality audits and evaluations of task outputs, define and refine quality benchmarks and scoring rubrics, analyze audit results to produce actionable insights and reports, and recommend process improvements to raise output quality.