yashnil mohanty

i’m a first-year at stanford. i like using machine learning on problems that matter, like health, climate, and making ai more reliable.

i bounce between research and building things. right now that’s virtual cell and sidequest.

Portrait of Yashnil Mohanty

things i’ve built

  1. virtual cell

    2026, ongoing

    zero-shot perturbation prediction

    can a model predict how a cell reacts to a gene knockdown if it has never seen that kind of cell? i’m testing this on crispri data from cancer cells, holding out one cell context at a time.

    python, pytorch, anndata

  2. sidequest

    2026, ongoing

    a travel planner

    ai finds things to do, and a scheduler makes them fit around routes, opening hours, weather, daylight, and transit. move one stop and only the plans that depend on it change.

    typescript, react, next.js

  3. chemcalculations

    2025–26

    a fast stand-in for fastchem

    fastchem works out which molecules show up in planet atmospheres. i trained a neural net on 4.8M of its runs, and it’s up to ~1,500× faster on a gpu. it’s on pypi.

    python, pytorch

  4. peak3

    2026

    basketball, measured at its peak

    basketball arguments, but with math. it scores every nba player-season since 1979–80 and finds each player’s best stretch. there are a few games built on top, too.

    python, next.js

experience

research

  1. 2025–26

    uc santa cruz · research intern, santa cruz, ca

    built a neural net that does exoplanet atmosphere chemistry much faster than the original code. it became chemcalculations.

  2. 2024–25

    national center for atmospheric research · research intern, boulder, co

    studied how wildfires throw off snowpack models. training on burned areas helped more than just telling the model how much burned. it became a first-author paper.

community

  1. 2025–26

    future speakers initiative · founder and president

    a nonprofit teaching public speaking and debate to underserved students. i built the curriculum and led a team of volunteer teachers. we reached 500+ students.

publications

  1. 2025 · journal

  2. 2025 · conference

  3. 2025 · conference

    Predicting Droughts: A Comparative Study of ARIMAX, LSTM, XGBoost, and Random Forest Models (opens in a new tab)

    Yashnil Mohanty · 11th International Conference on Computing and Artificial Intelligence (ICCAI), IEEE

  4. 2026 · software

awards

competitions

about

my work has jumped around: snowpack after wildfires, algal blooms, exoplanet chemistry, and now cells. the question i keep coming back to is when you can trust a model outside of what it was trained on.

lately i’m learning about ml, scientific computing, ai safety, computational biology, and climate.

for fun, i like baking, going for runs, reading, and watching sports. the best way to reach me is email.