Yash Shah

I am a Master's student in Computer Science, part of the School of Engineering at Stanford University. I matriculated in 2023 and am specializing in Artificial Intelligence. I currently work in Dr. Daniel L. K. Yamins' NeuroAI lab on topographic vision models, as well as with Dr. Ehsan Adeli on confounder-free systems.

Previously, I received my Bachelor's degree (Summa Cum Laude) from University of California, San Diego between 2019 and 2023, where I majored in Computer Science and minored in Humanities. There, I worked in Dr. Nadir Weibel's and Dr. Garrison Cottrell's labs and was a course assistant for CSE 151B under Dr. Garrison Cottrell (Fall 2022) and Dr. Rose Yu (Spring 2023).

Email  /  Google Scholar  /  GitHub  /  LinkedIn  /  Paintings  /  Quotes

Research

I'm broadly interested in computational neuroscience, developing models of the mammalian cortex that exhibit biologically-plausible emergent behavior.

Modeling Localized Synaptic Degeneration and Neural Plasticity in Visual Cortex

In this work, we simulate localized synaptic degeneration and neural plasticity using CORnet-S - a deep artificial, and recurrent, model of the brain.

Enhancing Accuracy, Time Spent, and Ubiquity in Critical Healthcare Delineation via Cross-Device Contouring

Here we explore contouring, a critical workflow aimed at identifying and segmenting tumors, to identify key behaviors, and later design a novel cross-device prototype that enables contouring on everyday touch devices.

Convolutional Neural Networks With Divisive Normalization Across Similar Features

In this paper, we explore using divisive normalization across neuronal sub-populations computing similar features to enforce local competition.

Industry Experience

A collection of my internship and full-time projects.

Software Engineer

Queried ~8M cards for various merchants using PostgreSQL and submitted for bulk network token enrollment. Analyzed 3-D secure transactions for 118 merchants and (1) enabled them to be processed through network tokenization first in production, (2) modified the way 3-D secure transactions are processed for certain processors and card types in the codebase so that future transactions can be directly processed through NT First.

Software Engineer Intern

Enabled network tokenization first for payment transaction processing for PayPal’s merchant partners to improve security and optimize authorization rates by around 40% for some merchants. Utilized data monitoring tools for monitoring code changes. Added a Foreign Tax ID Prompt to the Token Requestor ID (TRID) Onboarding UI to enable foreign merchants to apply for network tokenization.

Technology Development Intern

Developed a client portal using React and TypeScript and made Axios calls to the server using React Redux to onboard clients and allow them to place requests for Optum’s API services. Performed interface and architecture design research before building the portal for seamless user experience. Influenced company executives by presenting the business benefits of my project at Shark Tank presentations.

Other Technical Projects

These include coursework and side projects.

From Infant to Toddler to Preschooler: Analyzing Language Acquisition in Language Models

Here, we model the first six years of language acquisition by training an artificial language model on child-directed transcripts of spoken speech presented in an age-ordered fashion.

You Text, We Sketch: Text-guided Diffusion For Vectorized Sketch Generation

In this paper, we explore text-conditioned diffusion models for vectorized sketch generation by conditioning on text during training by feeding text tokens into a Transformer used as a text encoder, and introduce classifier-free guidance to improve sample quality.

Can a Set of Calibrated Weak Learners Create a Single Calibrated Strong Learner?

In this paper, we propose a unique way of implicitly calibrating AdaBoost (a traditional Machine Learning ensemble classifier) during training (as opposed to other post-hoc state-of-the-art techniques) through doubly-calibrating individual weak learners in the ensemble.

Image Captioning with Recurrent Neural Networks

In this work we present and analyze the working of a deep RNN in action by captioning images taken from the Microsoft COCO 2014 dataset. We explore various learning rate schedulers, ways of embedding and encoding, LSTMs, temperatures, etc.

PomoHero

A Pomodoro timer, based on the Pomodoro technique to help users better organize their schedules and get work done through time management.

Non-Technical Projects

From my Humanities minor.

Gender and Sexuality Through the Visual Imagery of Colors

In the stories of Colette and in Les Guérillères by Monique Wittig, we see different colors appear in different stories, bringing with them different symbolic connotations that may highlight themes within these stories that are not evident at first. Trying to interpret the visual imagery that colors bring might help us answer questions concerning why Colette and Wittig use them in the first place.

Thematization and Same-Sex Desire in “Song of the Ugly Maiden”

An interpretation of Eliza Cook’s poem “Song of the Ugly Maiden” is her yearning for a love for women. In this paper, I argue for this through analyzing thematization that occurs in the poem through rhyme scheme deviations, use of alliterations, word choices, and Cook having the Ugly Maiden sing her song as an interior monologue.

Women and Power in The Merchant of Venice

One of the questions that Shakespeare raises in his play The Merchant of Venice is the ability of women to have autonomy and exercise power. This autonomy and power, however, is confined to the domestic sphere between the husband and the wife. In the public sphere, women get to exercise power when they disguise themselves as men. Shakespeare, by portraying women in this manner, challenges the conventional social system of patriarchy.

The Plowman's Tale in The Canterbury Tales

An exercise in creative and imaginative writing, this paper presents a poem that the Plowman in Geoffrey Chaucer’s The Canterbury Tales could have sung if he were given a speaking role. A possible theme, which this paper presents, is the Plowman’s narration of the events of the Peasant’s Revolt of 1381 and an attempt to “quiet” Chaucer.

Du Bois, Racial Equality, and Gender Equality

In this argumentative essay, I view W. E. B. Du Bois’ text The Souls of Black Folk, being an undertaking for racial equality for Black Americans, through the lens of Simone de Beauvoir and her text The Second Sex, which has become, arguably, a classic text for the advocation of gender equality for women.

Honors and Awards

Over the years...

Computer Science and Engineering Department Excellence Award

for outstanding academic and leadership performance, and exceptional contributions to university life.

CSE Alumni Advisory Board Scholarship Award

for students who have demonstrated leadership and made contributions to the CSE community.

Christine Norris Award

for a student who best exemplifies what Humanities appreciated about Chris Norris: an enthusiasm for learning; the ability and desire to explore ideas and issues for their own sake; careful scholarship; and a concern for a clear and forceful articulation of ideas.

Revelle College Provost’s Honors Certificate

for achieving a minimum grade point average of 3.5 for at least 12 graded units for all three quarters in an academic year.

Revelle College Provost’s Honors Certificate

for achieving a minimum grade point average of 3.5 for at least 12 graded units for all three quarters in an academic year.

Outstanding Leadership Award

for exceptional efforts in guiding others and ensuring the success of “Pick-A-Pal” event.

Exceptional Diligence Award

for outstanding dedication to the programs executed in the 2019 Fall Quarter.

Design and source code from the Jekyll fork by Leonid Keselman of Jon Barron's website