studio

these are the things i've worked on.

01

tools

cool tools i've made

  • hikoki

    2026

    flight companion for your people

    texts the people you love when you take off, land, or run late. warm and personal, without another flight tracker to check.

    • track a flight or whole multi-leg trip
    • add parents, partners, and friends with messages written for each relationship
    • send gentle SMS updates on takeoff, landing, delays, and gate changes
    read the study →open hikoki ↗

  • hanga

    2025

    woodblock print deconstruction

    takes a finished image and works backwards to the blocks a printer would have cut.

    • upload an image, or open one of the bundled hiroshige prints
    • separate the keyblock outline from each colour block
    • read back the bokashi gradations and the hanmoto printing order
    read the study →open hanga ↗

  • carbonpaper

    2024

    ai document editor

    a real editor where the model suggests and you decide what lands in the document.

    • write and format in a document, not a chat window
    • select a passage and describe the edit you want
    • accept or reject each suggestion inline, with version history behind you
    read the study →open carbonpaper ↗

02

research

published and academic work.

  1. 01

    fig. 1 — oct scan and clinic metrics

    journal article · 2024

    A Novel Fusion Deep Learning Approach for Glaucoma Severity Classification – An Optical Coherence Tomography-Based Bimodal Artificial Intelligence Model

    Prasanna Venkatesh Ramesh, Puja Chinasigari, Shreyank Kadadi, Aditya Gupta, Archana Nivash, Murali Ariga, Pratheeba Devi Nivean, Sujatha Mohan

    TNOA Journal of Ophthalmic Science and Research, 62(2), 241–243

    one model reads the oct scan and another reads the numerical metrics from the same exam. their features are fused late, and the result grades severity as normal, mild to moderate, or severe rather than a yes or no.

    read the study →doi: 10.4103/tjosr.tjosr_23_24 ↗

  2. 02

    fig. 2 — the scored directed network

    thesis · 2023

    Novel Functional Interactomes of Proteins Across Brain Scales

    Shreyank Kadadi

    Undergraduate thesis, Computational and Systems Biology, UCLA

    defines a functional interactomes score for protein pairs, computes it from over 150 published experiments, and assembles the scored edges into a directed causal network that can be analysed topologically.

    faculty mentor: dr. sharmila venugopal, integrative biology and physiology

    read the study →thesis pdf ↗

  3. 03

    fig. 3 — one trial, 22 channels

    project paper · 2023

    EEG Classification using CNNs and RNNs

    Arvind Kalyan, Richard Jiang, Shreyank Kadadi

    Course project, UCLA

    four motor imagery classes from 22-channel eeg. pure recurrent models stayed near chance, but cnn/rnn hybrids reached 73% test accuracy, with dilated convolutions giving the largest single gain.

    read the study →code and paper ↗

03

game

games i've made

kokoro no tōkaidō

2025

platformer along the tōkaidō

five stations of the old road from edo to kyoto, playable in the browser.

celeste-style movement over public-domain hiroshige prints. each station teaches one mechanic and says nothing about it.

play it in the browser →open fullscreen ↗