David Gershony



Get Involved with My Work

I run a few research and engineering projects that I bring on collaborators for. These are real, in-progress efforts and a good way to ship meaningful work. If one fits your background, email me at david.gershony.7@gmail.com with your resume, the project you're interested in, and a few lines on why. I review every inquiry and reach out to those who are a strong fit.

Cognitive Similarity

Today's AI measures whether two pieces of text mean the same thing. We're hypothesizing a way to measure whether they feel the same to the human brain — using Meta's brain-response model (TRIBE v2) to capture things like mental effort and emotion that word-meaning alone misses. If it works, it's a new signal for content recommendation, readability, and understanding how media affects people. We've shown the brain signal diverges from standard embeddings; next we test whether the signal produces useful information, then ship a fast, distilled model anyone can use.

EigenLayer AVS Authority Scorer

Billions of dollars of Ethereum are staked on EigenLayer, yet there's no systematic way to know which operators and services are actually trustworthy — people currently go on yield, brand names, and gut feel. We treat the network as a graph and run a reputation algorithm so that an endorsement from a sophisticated operator counts more, surfacing the safest, highest-quality options before people put capital at risk.

Nomm

You trust your friends' taste more than a stranger's 4.2-star review. Nomm is a native iOS app that maps where your friends actually eat, turning your social circle into your restaurant guide — no anonymous reviews, no ads. Built in SwiftUI with MapKit and Supabase. The app is currently in beta testing on the App Store.


Last update: June 2026