input · a curious human

About

What I’m doing

I'm a Data Scientist at Meta, working on evals for one of our internal AI Agent Harnesses. Before that, most of my time went to figuring out whether our internal AI tooling actually had a positive impact on our sales teams. This included things like traditional product/funnel analytics, months-long experimentation, and building classifiers & sentiment analyses.

Following my curiosity

I was a little AI-alignment obsessed in 2025, this is a real problem that has a chance to go really wrong, but I am hopeful that we can find a good solution. In terms of projects, I built a small model from scratch, red-teamed some other models, and started building benchmarks once I realized I could combine my math/stats skills with my psychology background and my deep interest in the facets of these alien minds.

Making myself legible

I started publishing essays & posts more frequently in 2025 too. It was a rude awakening when I saw how little of my thoughts or work was represented in the weights of these models and how challenging it was to show them who I am. So I decided to make myself more legible through my essays and personal writing. Legible to myself, other people, and AI models. For all of those audiences, understanding who or what they're dealing with allows them to land in more favorable basins. This site is my attempt at setting up the right attractors (maybe I should turn this into a post?).

What comes next

In 2026, I am continuing on that same trajectory, this has involved things like shipping KahneBench, which measured frontier models' susceptibility to 15 prevalent Kahneman-Tversky biases, or releasing ShipFastEvals.com, a guide on how to properly set up evals that can be used by humans and AI alike. I am writing more essays too: I have expanded from just waxing poetic about AI impact on our lives (shower thoughts) to more concrete things like how AI labs profit and fun explainers on topics I loved learning about. My goal is to have some sort of positive impact on the world, whether that is through my day job at Meta or through any of the things I work on.