From Building to Proving by Ky'lin

Ky'lin's entry into Varsity Tutor's November 2026 scholarship contest

  • Rank: 1
  • 0 Votes
Ky'lin
Vote for my essay with a tweet!
Embed

From Building to Proving by Ky'lin - November 2026 Scholarship Essay

I want to get better at analysis.

For the last year I have been building an interpretable molecular descriptor framework called AVTSF. The work started as a question I could not shake: if a model predicts a quantum property of a molecule, can I trace exactly why? Most descriptors hand you a number. I wanted the number to be readable, something a chemist could check by hand against the structure.

That habit of reading backwards, from answer to reason, changed how I build. When the molecule builder on my chemistry platform Atomency kept confusing students about where a hydrogen bond could form, a 14-message exchange with Stanford psychologist Barbara Tversky taught me the problem was spatial, not computational. Students were not reading the model wrong. The model was drawn wrong. Rebuilding the builder around how people actually see molecules gave me the same lesson the AVTSF work did: a result you cannot inspect is a result you cannot trust.

Right now my analysis is good enough to pass a first peer review. My AVTSF manuscript was peer reviewed and accepted by the American Journal of Student Research, and it is not yet the kind of work I want to be doing in five years. In chemical engineering, a bad descriptor does not just lose points. It can send months of lab work down the wrong path. I want the discipline to formalize the instinct: statistical validation, error bounds, knowing when an interpretation is real and when it is wishful thinking.

Analysis is the skill that turns building things into proving things. I have spent a year building. I want the next four to prove.

Votes