Devan ShahI am a third year undergraduate in Computer Science at Princeton University pursuing a minor in mathematics. I am a member of the Hazan Lab led by Prof. Elad Hazan. Broadly, I am interested in post-transformer architectures, such as state space models, and applications of machine learning to understanding thought and behavioral patterns. I currently work on efficient attention mechanisms with the spectral transform unit. At Princeton, I cohost a reading group in natural language processing and chair our ACM chapter, with a general hope of increasing accessibility to research. Email / GitHub / Google Scholar / LinkedIn |
|
Work Experience |
|
SSM Research at the Hazan LabSeptember 2024 — PresentI work on accelerating inference with the Spectral Transformer, a state space architecture for language modeling produced by Google DeepMind Princeton in collaboration with the Hazan Lab. |
|
Recommendation Systems at TikTokMay 2024 — August 2024I improved the recommendation system for ecommerce videos for a projected $33,000,000+ increase in annual sales . I primarily focused on improving multi-interest modeling, training novel generative models (10B+ parameters) to better represent the interest space. |
|
EEG Vision Embeddings at CareYayaNovember 2023 — May 2024At CareYaya Health Technologies, I studied the reconstruction of viewed images from brain activity, improving performance by conditioning on additional user information. We aimed to better align EEG embeddings with CLIP space. |
|
Scooter Authentication at Cal Poly PomonaJune 2023 - April 2024At Cal Poly Pomona, I worked on continuously authenticating mobility scooter riders based on their posture patterns. We worked on embedding machine learning techniques for light-weight authentication and studied human motion in the search for a biometric identifier. |
|
Price Prediction at Ticket WalletJune 2023 — August 2023At Ticket Wallet, I worked closely with the CEO in an early-stage startup to advise and design data pipelines, pricing algorithms, and refine product pitches. Engineered the backend authentication and created a multi-modal ticket prediction system. |
Projects |
|
Wave Filtering for General Linear Dynamical SystemsDevan Shah* , Brandon Cho* (Equal Contribution) State Space Models paper |
|
ScooterId: Posture-Based Continuous User Identification From Mobility Scooter RidesDevan Shah , Ruoqi Huang, Nisha Vinayaga-Sureshkanth, Tingting Chen, Murtuza Jadliwala IEEE TMC 2024 paper |
|
Rider Posture-Based Continuous Authentication with Few-Shot Learning for Mobility Scooters (Student Abstract)Devan Shah , Ruoqi Huang, Tingting Chen, Murtuza Jadliwala AAAI 2024 paper |
|
Truly Adaptive Bloom FiltersDevan Shah* , David Yan* (Equal Contribution) ML Augmented Systems paper |
|
DreamScape: Denoising CLIP Embeddings with User Images for Improved Visualization ReconstructionDevan Shah CareYaya paper |
Surveys |
|
Discussion on "Why Does Deep and Cheap Learning Work So Well?"Devan Shah Physics of Deep Learning paper |
|
A Survey of State Space Models: From Linear Systems to LanguageDevan Shah* , Brandon Cho* (Equal Contribution) State Space Models paper |
|
Understanding Dynamic Algorithms for Packing-Covering LPs via Multiplicative Weight UpdatesDevan Shah* , Owen Yang*, Sunil Vittal* (Equal Contribution) Online Algorithms paper |
Coursework |
||||||||||||
Computer Science |
||||||||||||
|
||||||||||||
Theory |
||||||||||||
|
||||||||||||
Math |
||||||||||||
|
||||||||||||
Creative Writing |
||||||||||||
|