Kurtis Voris

Abstract

I build personalization systems for agentic experiences: user state and action models, recommender and search systems, with robust offline and efficient online evaluation methods to fuel personalized and proactive experiences.

Research & Product Interests

Sequential user modeling: Transformer architectures over interaction event streams for intent, state tracking, and next-action prediction
LLM fine tuning for user simulation: Enable realistic offline evals plus end-to-end agentic optimization, through cost-efficient SFT on real and synthetic data
Causal & statistical evaluation: Calibrated propensities, bias correction, causal inference, and efficient online experimentation for personalization
Agentic personalization: User representations and next-best-action policies that leverage user state to optimize long-term outcomes in conversational and proactive agent experiences

Opportunities of Interest

Consulting and contract engagements to scope and ship personalization, search, and fine tuned models into agentic AI systems
Principal full-time Applied Science / ML Scientist roles

QR code linking to this page — Scan to save