Research

AI Agents & Human–AI Decision-Making

I study when to delegate decisions to AI and how to evaluate and oversee its performance.

  1. Agents’ Last Exam
    2026 · NeurIPS (Evaluations & Datasets Track)

    Benchmarks AI agents on long-horizon professional tasks with verifiable outcomes.

    My contribution: Designed and built agent evaluation tasks for finance and supply chain applications.

  2. Probe What-If Worlds Before You Ship: Persona Simulation for Real-World Decision Making
    2026 · Under review

    Uses persona-based user simulation to compare product designs and study how user preferences shape AI assistant behavior.

    My contribution: Case development for the car assistant and home assistant studies.

  3. Learning to Delegate with Costly Audit
    2026 · Under review

    Studies how to learn when to delegate decisions and when to pay for verification.

Learning & Experimentation in Digital Platforms

I develop learning and experimental methods to improve decision-making under uncertainty on digital platforms.

  1. Best-Arm Identification with Generative Proxy
    2026 · Preprint

    Combines inexpensive model-generated proxy scores with real observations to identify the best option with fewer costly samples.

  2. Doubly Robust Switchback Experiments for Congested Online Platforms
    2026 · Under review

    Studies how to estimate treatment effects reliably in switchback experiments on congested online platforms.

  3. The Revenue Cost of Inference in Online Assortment Optimization
    2025 · Under review

    Studies the trade-off between revenue loss and the precision of revenue comparisons in online assortment experiments.