Applied Scientist Intern (Summer 2027)

At Lyft, our purpose is to serve and connect. We aim to achieve this by cultivating a work environment where all team members belong and have the opportunity to thrive. The Lyft Rider Science team is seeking an Applied Scientist intern to develop next-generation user simulation methods using state of the art AI methods. The goal of this project is to develop and validate LLM-based Rider Agents that can serve as behavioral proxies for real riders, and study when agent simulations can provide reliable signal about rider responses to product interventions before online experimentation. You will build agent-based simulation systems grounded in real rider context and behavioral data, evaluate their fidelity against observed rider behavior and historical experiments, and study where these simulations can accelerate product iteration and experimentation. This role combines LLM engineering, agent-based modeling, machine learning, and causal inference with direct applications to real-world rider products. The expected outcome is to build a working Rider Agent simulation prototype, establish an evaluation framework for measuring simulation fidelity and validate the framework using historical rider experiments.

Responsibilities:

Develop LLM-based Rider Agents that represent heterogeneous rider contexts, preferences, histories, and behaviors Build agent-based simulation environments for evaluating rider interactions with different product experiences and interventions Build evaluation pipelines to assess realism, robustness, and mechanism plausibility of simulated behavior against human data or established theory Analyze emergent behaviors and interaction dynamics in simulated populations under different user segment and marketplace conditions Conduct experiments and ablation studies on agent behavior, interaction dynamics, and simulation validity Apply the simulation framework to real Rider product problems and assess its usefulness for hypothesis generation, product iteration, and pre-experiment evaluation Communicate technical findings and recommendations to science, engineering, and product partners