Shubh Khanna

Email  /  Twitter  /  Linkedin  /  Scholar  

I'm a graduating M.S./B.S. student at Stanford studying computer science and math. Most of my prior work is in real-time visual decoding and in generative diffusion models.

I currently do diffusion research at Pika , and also do research with the Stanford Vision and Learning Lab (Fei-Fei Li) , as part of Stanford Artificial Intelligence Laboratory (SAIL).

Previously, I've built Ocular Diagnostics, a VC-backed brain health startup. We used eye movements to monitor the onset of neurobehavioral disorders.

In the past, I've:

• Developed a SOTA retrieval system (ICLR 2024) with the Stanford NLP Group
• Used pupil-tracking to build the most robust method for diagnosing neurobehavioral conditions (raised from Soma, published in Nature Scientific Reports)
• Made editing NeRFs go brrrr (2x speedup)
• Built first encoder-decoder architecture to generate text from MEG (neural data)

Some of my most fun side projects:

• Made a boba vending machine over a weekend and sold across Stanford (pictured above)
• Wrote Bayesian filtering algorithms, learned surgery, implanted sub-cortical devices into rodents (!), and decoded signals with these algorithms at Science
• Built a wearable that uses brain signals to steer a wheelchair
• Finetuned Mistral to make a personal email agent — email me to see :)
• Built tech for catching sex trafficking, building cheaper ventilators, and the NYC volunteering hub while running Coding for Impact

Publications

Illustrated features most pertinent to ADHD, built Naive Bayes classifier.

Real-time filtering algorithms and hough methods for pupil extraction.

ML pipelines to work with continuous time series data, segmenting eye features to diagnose ADHD.

Readings That
Inspire Me

© Shubh Khanna, 2023.