Principal Applied Scientist (User Modeling)

Accountabilities: Develop advanced user models: Design and build user world models that help AI agents understand where individuals are in their journey, infer their intentions, anticipate future needs, and predict the long-term consequences of different actions. Lead model post-training: Own end-to-end large language model post-training pipelines, including supervised fine-tuning (SFT), Direct Preference Optimization (DPO), and reinforcement fine-tuning (RFT/GRPO), running training workloads on GPU infrastructure. Build reward models: Develop and train efficient, multi-category process reward models (PRMs) that generate scalar reward signals from contrastive preference pairs to evaluate and improve agent behavior. Advance online evaluation: Design on-policy and online assessment frameworks that score live agent trajectories using LLM-as-a-Judge methodologies and other appropriate evaluation techniques. Translate evaluations into reward functions: Convert offline evaluation rubrics into generalizable reward functions capable of assessing diverse production traces and guiding continuous model improvement. Apply causal inference and reinforcement learning: Use sequential decision modeling, treatment-effect estimation, propensity scoring, and off-policy evaluation to understand user behavior and improve decision-making systems. Lead complex scientific initiatives: Independently define research problems, develop experimental approaches, analyze results, and deliver solutions from initial investigation through production implementation. Collaborate across disciplines: Partner with scientists, machine learning engineers, product managers, platform teams, and data specialists to align research with business needs and integrate new capabilities into production systems. Provide technical leadership: Mentor scientists and engineers, influence architectural and methodological decisions, communicate research findings, and establish high standards for experimentation and scientific rigor. Drive research innovation: Monitor advances in user modeling, reinforcement learning, causal inference, and generative modeling, translating promising techniques into measurable improvements in AI-powered consumer experiences. Requirements Education: PhD in Computer Science, Statistics, Economics, Machine Learning, or a related quantitative field, with strong expertise in reinforcement learning or causal inference. User modeling expertise: Hands-on experience modeling user behavior at scale, including sequence models built on event streams, sessions, or trajectories, and predicting user intent or long-term outcomes. Causal inference or reinforcement learning: Practical expertise in causal inference methods, such as propensity scoring, treatment-effect estimation, and off-policy evaluation, or experience applying reinforcement learning to real-world sequential decision problems. LLM training and evaluation: Experience with post-training methodologies, reward modeling, preference optimization, or evaluation frameworks for large language models and agentic systems is highly relevant to the role. Technical proficiency: Working proficiency in Python and machine learning frameworks such as PyTorch or TensorFlow, with the ability to run experiments independently and read, review, debug, and improve code. Scientific problem-solving: Demonstrated ability to independently scope and solve ambiguous, technically challenging problems, design rigorous experiments, and translate research findings into practical solutions. Technical leadership and mentorship: Proven ability to guide scientific or engineering initiatives, influence technical direction across teams, and support the development of other technical professionals. Preferred quantitative background: Knowledge of economics or econometrics, including demand modeling and consumer choice analysis. Research contributions: Publications or other demonstrated research contributions in reinforcement learning, causal inference, survival analysis, user modeling, or simulation. Generative modeling experience: Familiarity with generative sequence models or conditional language generation for behavioral or conversational simulation. Advanced modeling skills: Experience with rare-event modeling, probability calibration at scale, or other techniques for making reliable predictions in complex environments. Relevant industry exposure: Familiarity with two-sided marketplaces, real estate, or consumer journeys involving extended decision-making processes. Communication and collaboration: Strong analytical thinking, clear communication, and the ability to work effectively with cross-functional teams to turn research objectives into production impact. Benefits Competitive annual base salary: $195,600–$312,400 in California, Connecticut, Maryland, Massachusetts, New Jersey, New York, Washington State, and Washington, D.C. $185,800–$296,800 in Colorado, Hawaii, Illinois, Maine, Minnesota, Nevada, Ohio, Rhode Island, Vermont, and Virginia. Compensation varies by location and is determined by factors such as experience and performance. Salary ranges for other locations may differ. Equity compensation: Eligibility for equity awards in addition to base salary, with actual awards dependent on experience, performance, and location. Remote work flexibility: Remote position for eligible employees based in the United States, subject to limited location restrictions. High-impact research opportunities: Work on advanced AI systems designed to improve personalized consumer experiences at significant scale. Cutting-edge technology: Opportunities to explore reinforcement learning, causal inference, LLM post-training, reward modeling, and advanced user simulation. Professional growth: Opportunities to influence scientific strategy, lead complex research initiatives, and mentor other scientists and engineers. Collaborative environment: Work with multidisciplinary teams across research, product, engineering, platform, and data functions. Inclusive workplace: A commitment to equal employment opportunity and a culture that values diverse perspectives and professional development. Accessibility support: Reasonable accommodations are available for qualified applicants and employees who require them.