About

I work on AI agents at Google Cloud AI, focusing on how to evaluate them and make them faster, more reliable, and less expensive to run.

Email me GitHub LinkedIn

How far ahead can we predict a chaotic system?

In the Hénon–Heiles model, a particle can escape through one of three exits. I tested whether four noisy position measurements could predict escape when both the energy and starting region were new to the predictor.

The short history helped at nearby forecast times, but the gain faded further ahead. Adding a measure of instability did not reliably improve the predictions.

Read PDF 6 pages · 713 KB

Three simulated Hénon–Heiles paths at energy 0.18 wind through the central region before exiting upward in blue, lower-right in gold, and lower-left in coral. Circles mark their starting positions, and faint curves mark potential-energy contours. These are illustrative trajectories from the study, not a chart of prediction accuracy. — Three example trajectories from the study, each at the same energy and reaching a different exit. Faint lines show potential-energy contours. The paper tests how well short observations predict escape.

Strust

Rewriting a program is easier than knowing whether it still behaves the same. I built Strust to run COBOL programs and their Java replacements on shared inputs and compare the results.

Read PDF 2 pages · 97 KB

all research