Skip to main content
CareerApp
Zillow

Principal Applied Scientist, Generative AI

Zillow

Remote-USA · Remote · full time · Lead / Staff / Principal

$181,800 – $305,700

Listed on Zillow’s own careers site. You apply with them directly — we never stand between you and the employer.

What this role is

Lead the design and deployment of reinforcement learning systems that align large language models with user value and safety objectives at Zillow. This principal-level role suits experienced applied scientists who can own end-to-end post-training strategies and mentor others while shipping models into production.

Our summary, not Zillow’s wording. The full posting is on their site.

Skills this role names

Log in to see which of these are already on your profile.

What they ask for

Required

  • PhD or equivalent in Computer Science, Electrical Engineering, Statistics, or related field
  • Expertise in post-training techniques including DPO, RLHF, RLAIF, and preference modeling
  • Experience evaluating and monitoring aligned models with win-rate experiments and safety metrics
  • Strong understanding of modern transformer-based LLMs and orchestration frameworks
  • Experience working cross-functionally with engineering, product, and compliance teams
  • Technical leadership and mentorship capabilities
  • Ability to communicate complex technical ideas to both expert and non-expert audiences

Nice to have

  • Experience in high-stakes domains such as marketplaces, finance, healthcare, or regulated verticals
  • Background in bandits or other reinforcement learning applications
  • Open-source contributions or external thought leadership through publications or talks
  • Non-traditional background with transferable skills

Turn on analytics and we load Google Analytics: Google gets the pages you open and what you do here — searches, jobs you view, jobs you apply to — and sets its own cookies. Leave it off and the only cookies we set are your login, your theme, and this answer. Accept All also records a yes to advertising, which nothing uses yet. Privacy Policy.