Principal Applied Scientist – User Modeling

This is a fully remote position, open to applicants in United States.

📋 Description

• Develop user world models that enable agents to comprehend each shopper’s journey, objectives, and the impact of actions on outcomes.

• Manage LLM post-training pipelines, including SFT, DPO, and reinforcement fine-tuning (RFT/GRPO), end-to-end utilizing GPU infrastructure.

• Create and train multi-category reward models (PRMs) that generate scalar signals from contrastive preference pairs.

• Design on-policy and online assessments employing LLM-as-a-Judge frameworks to evaluate live agent trajectories.

• Convert offline evaluation criteria into generalizable reward functions applicable to arbitrary production traces.

• Lead complex and ambiguous projects from start to finish in collaboration with product, engineering, science, platform, and data teams.

• Provide technical guidance and mentorship.

• Transform cutting-edge AI research into impactful production applications for Zillow’s home-shopping experiences.

⛳️ Requirements

• PhD in Computer Science, Statistics, Economics, Machine Learning, or a related quantitative discipline, with a focus on reinforcement learning or causal inference.

• Practical experience modeling user behavior at scale using sequence models over event streams, sessions, or trajectories.

• Proven experience in predicting user intent or long-term outcomes.

• Strong understanding of causal inference, including propensity scoring, treatment-effect estimation, and off-policy evaluation, or reinforcement learning applied to real sequential decision-making problems.

• Capability to independently define and solve ambiguous problems from start to finish.

• Proficient in providing technical leadership to scientists and MLEs.

• Working knowledge of Python and machine learning frameworks like PyTorch or TensorFlow.

• Ability to conduct experiments and read, review, and debug code.

• Background in economics or econometrics, such as demand modeling or consumer choice.

• Published research in RL, causal inference, survival analysis, or user modeling/simulation.

• Familiarity with generative sequence models or conditional language generation for behavioral or conversational simulation.

• Experience in rare-event modeling and probability calibration at scale.

• Knowledge of two-sided marketplaces, real estate, or long-horizon consumer journeys.

🏝️ Benefits

• Equity awards based on factors such as experience, performance, and location.

• Remote work from a preferred physical location within eligible U.S. locations.

• Flexible work arrangements through Cloud HQ.

• Equal employment opportunity and accommodation support.