Hi, I’m Kyle.

I work as a machine learning engineer at Melange building automated patent search. Before that I studied math at UT Austin and Williams College. In the past I’ve worked on neural network interpretability, prediction markets, and NBA analytics. I’m from Austin, Texas.

Selected Projects

  • Steering results across models and tasks

    Post-hoc reasoning in chain of thought

    LLMs often commit to an answer before chain-of-thought, then rationalize it. We can detect these answers with linear probes and switch them via steering.

    Paper · Post · LessWrong

  • Uncertainty metrics across datasets

    Mapping from Meaning

    Improving the uncertainty calibration of LLMs by sampling over semantically equivalent perturbations of the same prompt. [AAAI 2025]

    Paper

  • NBA season forecast table

    NBA Forecast

    Massey-style team ratings, player plus-minus ratings, game predictions, Monte Carlo season simulations, and player career projections.

    Forecast

Feel free to reach out! My email is
{firstname}{lastname}2000 at gmail dot com.