Research Scientist 4 - Machine Learning and Inference Research, LLM Post-training

Netflix Netflix · Big Tech · New York, NY +3 · Data & Insights

Research Scientist 4 at Netflix focused on post-training LLMs, particularly using RL techniques, and potentially other areas like reasoning, alignment, distillation, tool use, memory, and calibration. The role involves fundamental research, publishing at top venues, and translating research into impact at scale within the consumer domain.

What you'd actually do

  1. Define and execute a strong research agenda with both internal and external visibility
  2. Disseminate knowledge effectively and inspire others
  3. Collaborate with colleagues to deliver tangible impact
  4. Help foster an open environment of innovation, intellectual rigor, and curiosity
  5. Leverage your technical expertise to shape roadmaps, collaborate across functions, and bring new ideas from exploration to impact

Skills

Required

  • Ph.D. in Computer Science or related field
  • Specialization in post-training LLMs
  • Expertise in RL (e.g., RLVR, RLHF, offline or online, policy- or value-based)
  • Experience with reasoning, alignment, distillation/compression, tool use, memory, calibration
  • Track record of top-tier publications
  • Strong technical communication skills
  • Self-motivated
  • Curiosity and judgment

Nice to have

  • Passion for collaboration
  • Eagerness to elevate the broader organization

What the JD emphasized

  • Ph.D. in Computer Science or a related field with a specialization in post-training LLMs for downstream tasks, especially using RL (e.g., RLVR, RLHF, offline or online, policy- or value-based), and possibly also including reasoning, alignment, distillation/compression, tool use, memory, calibration, or related.
  • A track record of top-tier publications demonstrating deep expertise in the specialization.

Other signals

  • post-training LLMs
  • RL (e.g., RLVR, RLHF, offline or online, policy- or value-based)
  • reasoning, alignment, distillation/compression, tool use, memory, calibration